Where We Left Off
This is an advocacy document. It is the case for the prosecution against Congress governments in power. The selection of material is one-sided on purpose. The evidence in it is not. Every serious claim carries a grade showing how well it is proved, the other side is given its strongest argument, and where the other side wins, this book says so. A second series applying the same test to the government of 2014 onwards is planned.
Part Fourteen ended in May 1996, with the Congress reduced to a hundred and forty seats and out of office. What happened next is not this book’s subject and can be given in a paragraph.
Between 1996 and 2004 India had six Prime Ministers and four governments. The Congress was out of power for the whole of it. The Bharatiya Janata Party, which had a hundred and twenty seats in 1991, formed governments in 1996, 1998 and 1999, and Atal Bihari Vajpayee held office from 1998 until 2004. That period belongs to the second series this book has promised, not to this one.
In May 2004 the Congress came back. This part covers the five years that followed — the first Manmohan Singh government, usually called UPA I.
What is in this part and what is not
The four subjects agreed for this part are the terror wave, the Mumbai attacks of November 2008, the cash-for-votes affair and the nuclear deal. They are all here.
What is not here is the money. The 2G spectrum allocation, the Commonwealth Games, Adarsh and Coalgate are the subject of Part Sixteen, and this part does not touch them even where they begin inside these five years — the spectrum allocation of January 2008 is the clearest case. Part Eleven did the same thing when it declined to re-describe the five days that Part Ten had already covered. Splitting a subject and then telling it twice is how a series loses a reader.
Every account of these five years assumes that the Prime Minister ran the government. It is such a basic assumption that nobody states it, which is exactly the kind this series looks for.
Consider the arrangement. Manmohan Singh was not the leader of his party — Sonia Gandhi was. He had not won a Lok Sabha seat; he sat in the Rajya Sabha, elected by a state legislature. He had no political base of his own and had never had one. The party’s organisation, its candidate selection and its money answered to somebody else.
This produces a genuine problem for a book like this one. When something in these five years went wrong, the question “who decided this?” may have no locatable answer at all — not because it is hidden, but because the office had been separated from the authority.
Part Eight found that Indira Gandhi’s power came from an office that had never had its checks tested. Part Twelve found the same office producing identically shaped failures under a completely different occupant. Part Fifteen finds a third variation: the office occupied by somebody who did not hold the power that goes with it. Three governments, three arrangements, and the finding survives all three, which is why this series argues that the problem is structural rather than personal.
This part therefore takes more care than usual about who is being charged. Where an act was the central government’s, it says so. Where it was a state government’s — and several of the worst things here were — it says that instead, and names the state and the party running it.
1The Government of the Third Choice
The 2004 election was expected to be a formality. The governing alliance was campaigning on a slogan about how well the country was doing. Almost every forecast, including most of those inside the Congress, expected it to win.
It did not. The Congress won a hundred and forty-five seats, more than any other single party, and put together a coalition — the United Progressive Alliance — with support from outside by the Communist parties, who had around sixty seats and did not join the government.
Then something happened that has no parallel in the period this series covers.
1.1 — The result
The governing alliance went into the 2004 election on a slogan about how well the country was doing. Growth was above eight per cent. The stock market was high. The campaign was built around that.
It lost. Not narrowly, and not for a reason anybody had forecast.
Growth of eight per cent is a national average. It is made by adding what happened to a software engineer in Bangalore to what happened to a landless labourer in Bihar and dividing by two.
Between 2004 and 2009 the number of Indians employed in the software and services industries that produced most of that growth was around two million, in a workforce of roughly four hundred and seventy million.
So the campaign told nine people in ten that they were doing well, on the strength of what was happening to the tenth. They knew whether they were doing well. They did not need to be told, and they resented being told wrong.
This is the same finding as Part Five’s, in a new setting: a national number can be true and describe almost nobody.
That gap between the average and the experience is the whole result, and it can be checked against where the seats actually moved.
The claim that the campaign misread the country is easy to make afterwards and hard to prove. The evidence for it is in where the seats moved.
The governing alliance’s losses were concentrated in states where the previous years had been hardest in the countryside, and its strongest results were urban. In Andhra Pradesh, where a government closely identified with technology-led growth had governed for nine years, it was removed almost completely.
An election is the only instrument in this series that measures what people think about their own circumstances rather than what a statistic says about them, and in 2004 the two disagreed.
The Congress won a hundred and forty-five seats, more than any other single party. It put together a coalition — the United Progressive Alliance — with support from outside by the Communist parties, who had around sixty seats and did not join the government.
The consequence for this part is that the government arrived with a specific instruction from the electorate, and it knew what the instruction was. Chapter Two is what it did with it.
1.2 — The refusal
Sonia Gandhi, who had led the party to the win and was its president, declined to become Prime Minister. She proposed Manmohan Singh instead. He was sworn in on the twenty-second of May 2004.
This book will not pretend the refusal was uncomplicated. There was an active campaign against her holding the office on the ground that she had been born in Italy, and a constitutional challenge was threatened. Whether the decision was principle, calculation, or both is not established and Chapter Ten lists it among the things not known.
What matters here is not the motive but the arrangement it created, because that arrangement is the subject of this whole part.
A coalition is several parties agreeing to govern together, sharing ministries.
Support from outside is different and it matters here. A party promises to vote with the government but takes no ministries and holds no office. It gets influence without responsibility, and it can walk away at any moment without resigning from anything.
The Left parties supported UPA I from outside for four years. In July 2008 they walked away. Chapters Seven and Eight are about what happened next.
1.3 — What the arrangement actually was
Set out plainly, the position from May 2004 was this. The Prime Minister was an economist who had been Finance Minister in the government of Part Fourteen, had never won a Lok Sabha election, and sat in the upper house. The president of his party, who commanded the loyalty of its MPs, held no government office. The government’s majority depended on parties that were not in it.
Three separate people or bodies held pieces of what is normally one job.
Imagine a company where the chief executive does not own shares, cannot hire or fire the senior staff, and answers to a chairman who is not on the payroll — while a third party that owns no shares can shut the business down by writing a letter.
Nobody would run a company that way. India was run that way for five years, and then for five more.
It is worth saying that it worked, in the sense that the government lasted its term and won the next election. Whether it worked in the sense that matters — that somebody could be held responsible for a decision — is the question of this part.
An additional body was created that has no clear precedent. The National Advisory Council, chaired by Sonia Gandhi, sat outside the government and advised on legislation. Several of the laws in Chapter Two came through it.
The defence says: this is a description of ordinary coalition government, dressed up as something sinister. Parliamentary systems have always separated party leadership from government office — a party president who is not Prime Minister is normal in most democracies. The Left’s outside support was a public arrangement with a written common minimum programme, not a secret. And the National Advisory Council was an advisory body whose recommendations Parliament was free to reject, which it sometimes did.
The reply: each of those points is true separately. The difficulty is all three at once. A Prime Minister who lacks a party base can be overruled by his party; a party president outside government cannot be questioned in Parliament; and an advisory council answers to nobody at all. The result is not a conspiracy. It is that for any given decision in these five years, the number of people who can be asked about it in the House is smaller than it should be.
Where the defence wins: on results. Chapter Two is the answer, and it is a serious one.
1.4 — What was promised
The alliance published a Common Minimum Programme — a written list of what it would do, agreed with the parties supporting it from outside. This is unusual and it is useful, because it means the government can be measured against a document it wrote itself rather than against what its critics say it should have done.
Part One set the rule: the strongest charge against a government is one built from its own paper. The Common Minimum Programme promised, among other things, a law guaranteeing employment in rural areas, a right-to-information law, a large increase in spending on education and health, and — this one matters for Chapter Nine — a law against communal violence.
| Promised, May 2004 | What happened by May 2009 |
|---|---|
| A right-to-information law | Delivered. Act passed June 2005, fully in force 12 October 2005. Chapter Two. |
| Guaranteed rural employment | Delivered. Act passed 2005, rolled out from February 2006. Chapter Two. |
| Recognition of forest-dwellers’ rights | Delivered. Forest Rights Act 2006 — see Part Thirteen, which covers what it did and did not do. |
| 6% of national income on education | Not delivered. Spending rose but did not approach the figure. |
| 2–3% of national income on health | Not delivered. Public health spending stayed close to 1%. |
| A law against communal violence | Not delivered. Drafted in 2005 and repeatedly afterwards. Never enacted, in ten years of office. Chapter Nine. |
Part Eleven recorded the communal violence bill as drafted in 2005 and never passed. This part is where that promise came due, and it is a debt this book has been carrying for four parts.
From May 2004, three different people or bodies held pieces of one job: a Prime Minister with no party base, a party president with no government office, and allies with a veto and no responsibility.
The government wrote down what it would do. It kept three of those promises and broke three. The one it broke that costs most is the law against communal violence — and Chapter Nine explains why that particular failure runs through half this part.
2What Was Built
Part Fourteen put its credit chapter second, before the charge, and explained why: placed at the end it would read as a sweetener produced late. The same rule applies here and for the same reason. What follows is the strongest thing UPA I did and it goes in first.
2.1 — The Right to Information Act
Before 2005, a citizen who wanted to know something a government department knew had no way to make it tell him. There was a Freedom of Information Act passed in 2002 that had never been brought into force in any useful way.
The Right to Information Act, 2005, changed that. It passed Parliament on the fifteenth of June 2005 and came fully into force on the twelfth of October.
Any citizen may ask any public authority for information it holds. The authority must answer within thirty days — within forty-eight hours if the question concerns somebody’s life or liberty.
If it refuses without proper reason, an Information Commission can fine the official personally. That last part is the whole law: without a penalty on a named person, a duty to disclose is a suggestion.
The scale of what followed is easy to state and hard to take in.
Roughly forty-eight hundred applications a day, every day, for twenty years.
Most of them are not journalism. They are a woman asking why her widow’s pension has not arrived, a labourer asking what wage was recorded against his name, a family asking why a ration card was refused.
Before 2005 the answer to all of those was: you have no right to ask. That is the sentence the Act deleted, and it is the largest single transfer of power to ordinary people in this series.
This book is a prosecution document and it has to be careful with an achievement, so here is the honest boundary of the claim. The Act came from a movement outside government — a campaign in Rajasthan that had been running for a decade, and state laws in nine states before the central one. The UPA did not invent it. What the UPA did was pass it in a strong form, with the penalty clause intact, when a weaker version would have been easier and would have been accepted.
The test of whether a transparency law has teeth is what later governments try to do to it.
The 2002 Act it replaced was never brought properly into force by the government that passed it. The 2005 Act has since been amended, in 2019, to change the tenure and status of the Information Commissioners — the officials who impose the penalties — and that amendment was opposed on the ground that it weakened their independence.
Nobody bothers to weaken a law that does nothing. The attempts to alter it are the best available evidence that it worked.
2.2 — A hundred days of work
The National Rural Employment Guarantee Act passed in 2005 and began operating in two hundred districts in February 2006, extending to the whole country by 2008.
It does something no earlier Indian anti-poverty programme did. It does not offer work. It creates a legal right to it: any rural household whose adults want unskilled manual work is entitled to a hundred days of it in a year at a stated wage, and if the work is not provided within fifteen days, the household is entitled to be paid anyway.
A scheme is money a government sets aside to do something. When the money runs out, the scheme stops. Nobody can sue.
A right is an entitlement a named person holds. If it is not honoured, that person can go to a court and the government must find the money.
Part Thirteen made this distinction about tribal land and found that a right is worth the independence of whoever enforces it. The employment guarantee is the counter-example in this series: a right where the enforcement mechanism was built into the same law.
Its record is genuinely mixed and this book gives both halves. Studies found real effects on rural wages and on distress migration, particularly in Rajasthan and Andhra Pradesh. They also found large-scale leakage — false muster rolls, wages recorded against people who never worked, delays in payment running to months.
What is worth noticing is how the leakage came to be known.
What is worth noticing is that an anti-corruption measure and a welfare measure are usually treated as different kinds of thing — one polices, the other spends. These two were not. The employment law was written so that its own records had to be published and read aloud in the village, and the transparency law passed four months earlier made those records demandable by anybody.
So the corruption in the employment programme was documented, in enormous detail, mostly by the people it was stealing from, using a law the same government had passed the same year. That is why this part knows so much about how badly the programme leaked. The scandal is legible because the government made it legible — which is the opposite of Part Thirteen, where nobody had ever added up the displaced, and of Part Five, where no beneficiary of freight equalisation ever saw the pipe.
2.3 — What these laws did not do
Both laws have a boundary and it is the same boundary in each case: they give a citizen a way to compel an answer, and they do nothing about what happens after the answer arrives.
The transparency law lets somebody establish that a wage was recorded against his name and never paid. It does not get him the wage. To do that he must go to a court, which costs money he does not have and time he cannot take. The employment law lets him claim unemployment allowance if work is not given within fifteen days. In practice the allowance is paid by the same state government that failed to provide the work, out of its own funds, and it is very rarely paid.
Studies of the employment programme in its early years found delays in wage payment routinely running to weeks and often months, against a statutory requirement of fifteen days.
For a household earning at that wage, a two-month delay is not an inconvenience. It is the difference between eating and borrowing at the rates a landless family gets offered, which in much of rural India means five to ten per cent a month.
A legal right that arrives two months late has, for the person holding it, already failed — and the law provides no penalty on anybody for the delay.
The other two Common Minimum Programme promises were not kept at all, and they are large.
| Promised in 2004 | Where it stood by 2009 | Why it matters |
|---|---|---|
| 6% of national income on education | Rose, but nowhere near the figure | The target dates from 1968 and has been promised by every government since |
| 2–3% of national income on health | Public health spending stayed close to 1% | India’s out-of-pocket health spending is among the highest in the world; the gap is paid by households |
These are the two commitments in Table 1 that required sustained money rather than a single Act. That is the pattern: what could be done by passing a law was done, and what required a permanent claim on the budget was not.
And the Sachar Committee, which reported in November 2006 and established the position of Indian Muslims in detail, produced recommendations. An Equal Opportunity Commission was proposed and never created. A national data bank on the position of communities was proposed and never built.
Part Nine found that a correct diagnosis in an organisation with no route for it to travel upwards goes nowhere. The Sachar report is the clearest case of that in this part: the measurement was made, honestly and publicly, and nothing was built to act on it.
2.4 — The rest of the column
Three more, stated briefly.
The Forest Rights Act, 2006, whose preamble admits that the rights of forest dwellers were never recorded and calls this a historical injustice. Part Thirteen covers it in full, including what it failed to fix.
The Sachar Committee, appointed in March 2005 under a former Chief Justice of the Delhi High Court, reported in November 2006 on the condition of Indian Muslims. It found them substantially worse off than the national average on income, education, public employment and access to credit. It is the first serious official measurement of the question.
And in January 2006 the government announced a package of seven hundred and seventeen crore rupees for the victims of November 1984. Part Eleven set that out in detail, including the fact that around two hundred crore of it was never distributed because of disputes over who was entitled. It goes in this column, with that qualification attached.
Strong, and stronger than this book expected. Two laws that changed what a citizen can demand from the state, one that recognised rights never recorded in a century, and the first honest measurement of a minority’s position.
The qualification: the ideas came from outside government — from campaigns, from earlier state laws, from committees. What this government did was pass them, in strong forms, when weak ones were available.
Passing somebody else’s good idea, without watering it down, is a real thing to have done, and it is rarer in this series than inventing one.
Now the other side of the ledger, which occupies six of the remaining eight chapters.
UPA I passed a law letting any citizen force any department to answer a question, and a law making rural work an enforceable right rather than a scheme.
Roughly forty-eight hundred information requests a day have been filed since. The 2019 attempt to alter the law is the best evidence that it bites.
And the most useful thing about it: the corruption in the employment programme is known in detail because this government built the tool that exposed it. Nothing in the previous thirteen parts looks like that.
3The Wave
Between 2005 and 2008 India was bombed roughly once every three months. That sentence is easy to write and almost impossible to hold in the head, so this chapter puts it in a table first and then explains what the table does not show.
| Date | Place | Killed | Who was eventually held responsible |
|---|---|---|---|
| 29 Oct 2005 | Delhi, three markets before Diwali | ~60 | Prosecutions collapsed; two acquitted, one convicted then acquitted on appeal |
| 7 Mar 2006 | Varanasi, temple and railway station | ~20 | Convictions in 2022, sixteen years later |
| 11 Jul 2006 | Mumbai, seven suburban trains in eleven minutes | ~189 | Twelve convicted 2015; all acquitted by the Bombay High Court in 2025 |
| 8 Sep 2006 | Malegaon, near a mosque and graveyard | ~37 | Nine Muslim men arrested — wrongly. Chapter Four |
| 18 Feb 2007 | Samjhauta Express, the train to Lahore | 68 | Four accused; all acquitted 2019 |
| 18 May 2007 | Mecca Masjid, Hyderabad, at Friday prayers | 9, plus 5 shot by police afterwards | Muslim men arrested first, released; five later accused, all acquitted 2018 |
| 11 Oct 2007 | Ajmer, the Sufi shrine, during Ramzan | 3 | Two convicted 2017; the principal accused acquitted |
| 13 May 2008 | Jaipur, nine blasts | ~63 | Four sentenced to death 2019; all acquitted by the Rajasthan High Court 2023 |
| 25–26 Jul 2008 | Bangalore, then Ahmedabad | ~57 | Ahmedabad: 38 sentenced to death 2022, appeals pending |
| 13 Sep 2008 | Delhi, five blasts | ~30 | Prosecutions continuing |
| 29 Sep 2008 | Malegaon again, and Modasa | ~7 | Seven accused; all acquitted 2025 |
| 26–29 Nov 2008 | Mumbai | 166 | One attacker captured and convicted; nine killed. Chapter Five |
Casualty figures vary between sources by ten to twenty per cent in most of these cases and the numbers above are the commonly cited ones. Where a figure is contested it is given with a tilde. The final column is the position as at the date of writing and several of these cases remain live; Chapter Ten deals with what that means.
3.1 — What the table shows
Read down the last column rather than the third. That is where this chapter is.
Around six hundred people were killed in those attacks. For the great majority of those deaths, nobody is serving a sentence today. Convictions were obtained and then reversed on appeal. Men were held for a decade and then acquitted. In several cases one set of people was arrested, released, and a completely different set arrested afterwards.
Take the Mumbai train bombings of July 2006. A hundred and eighty-nine people were killed in eleven minutes on seven commuter trains.
Twelve men were convicted in 2015 — nine years later. Five were sentenced to death. In 2025 the Bombay High Court acquitted all of them, holding that the prosecution had failed to prove its case.
Those men spent around nineteen years in prison. One died there. If the High Court is right, the state took nineteen years from twelve men and the actual bombers have never been identified. If the High Court is wrong, twelve mass murderers have just been released.
There is no version of that sentence in which the Indian state comes out of it well.
3.2 — Eleven minutes in July 2006
One of those rows deserves its own section, because it is the largest attack of the period apart from 26/11 and because of what happened to it afterwards.
On the evening of the eleventh of July 2006, seven bombs went off in the first-class compartments of seven suburban trains on the Western line in Mumbai, over about eleven minutes, during the evening rush. Around a hundred and eighty-nine people were killed and more than eight hundred injured.
Thirteen men were charged under MCOCA, the Maharashtra organised-crime statute. The trial ran for eight years. In 2015 twelve were convicted — five sentenced to death, seven to life. One had been acquitted.
In July 2025 the Bombay High Court acquitted all twelve, holding that the prosecution had failed to prove its case.
The reasons the High Court gave are the reasons that recur in this part, and they are not exotic.
The confessions were made to police officers, admissible only because MCOCA suspends the ordinary rule, and the court found them unreliable. Identification of the accused by witnesses came years after the event. The recovery of explosives and the chain of custody of the material evidence did not hold up.
None of that was discovered in 2025. All of it was on the record from the beginning, which is what an acquittal on the same evidence nineteen years later means.
What that leaves is a hundred and eighty-nine deaths in a commuter rush hour, in India’s largest city, in 2006, for which no conviction now stands and no other suspects have ever been charged.
3.3 — The law that was repealed and then rewritten
One of the first things UPA I did was keep a promise. One of the last things it did was take it back, and the two acts bracket this chapter.
The Prevention of Terrorism Act had been passed in 2002 by the previous government. It allowed detention for up to six months without charge, made confessions to police officers admissible, and reversed the ordinary presumption on bail. The Common Minimum Programme promised to repeal it, on the ground that it had been used against political opponents and minorities rather than against terrorists. It was repealed in September 2004.
In December 2008, in the five weeks after the Mumbai attacks, Parliament amended the Unlawful Activities (Prevention) Act.
| POTA 2002, repealed Sept 2004 | UAPA as amended, Dec 2008 |
|---|---|
| Detention without charge up to 180 days | Detention without charge up to 180 days |
| Police custody up to 30 days | Police custody up to 30 days |
| Bail effectively barred where the court sees a prima facie case | Bail effectively barred where the court sees a prima facie case |
| Confessions to police officers admissible | Not restored. This is the one real difference. |
| Sunset clause: lapsed after three years unless renewed | No sunset clause. Permanent. |
The confession provision was the single worst thing in POTA and it was not brought back — which matters, and Chapter Four shows why. But it survives in state statutes like MCOCA, which the central government did not touch.
So the promise was kept in 2004 and substantially unmade in 2008, minus one provision and minus the expiry date. Part Nine found the identical move: the instruments used to seize a moment were repaired after the Emergency and the instruments used to hold the country were not, and the National Security Act was passed in 1980 by the Congress government that returned.
The defence says: a hundred and sixty-six people had just been killed by men who arrived by sea and were directed by telephone from another country. A state that does not change its law after that is not principled, it is inert. And the change was made through Parliament, in public, with the opposition supporting it — this was not an ordinance.
The reply: the objection is not that the law was strengthened. It is that the government had argued for four years that these specific powers were dangerous because they get used against the wrong people — and then, four weeks after a catastrophe, enacted them permanently and without an expiry date. Chapter Four is what “used against the wrong people” looks like, and it happened while POTA was repealed. The problem was never the statute’s name.
3.4 — What the table does not show
The bombings of 2006 and 2007 divide into two groups, and the division is the most important fact in this part.
One group struck markets, trains and city centres. The other struck a mosque during Friday prayers, a Muslim graveyard on a night of remembrance, a Sufi shrine during Ramzan, and a train to Pakistan carrying mostly Pakistani passengers.
At the time, all of them were investigated on the assumption that they had been done by Islamist groups. In the second set, that assumption was wrong.
The reversal did not come from journalism or from a campaign. It came from inside the state, in two steps.
First, in 2008, the Maharashtra Anti-Terrorism Squad under Hemant Karkare investigated the second Malegaon bombing and arrested members of a Hindu organisation, including a serving Lieutenant Colonel of the Indian Army. That was the state arresting people its own earlier investigations had not been looking at.
Second, in December 2010, Swami Aseemanand made a confession before a magistrate — the strongest form of confession in Indian law, because a magistrate must first warn the maker and satisfy himself it is voluntary — stating that Hindu extremist groups had carried out the Malegaon, Mecca Masjid, Ajmer and Samjhauta attacks.
He later retracted it, and the cases built on it have since collapsed. But by then the men originally arrested had been in jail for years, and a separate agency of the same state had filed charge sheets saying they were innocent.
Everything in Chapter Four follows from that.
Every account of a terror wave assumes that the measure of it is the number of people killed. That is what headlines count, what memorials record, and what this chapter’s table puts in the third column.
But look at the fourth column instead. Across twelve attacks and roughly six hundred deaths, the thing that varies most is not casualties. It is whether the state ever correctly identified anybody.
Counting the dead measures what the attacker did. It does not measure anything the government did, which means a book that judges governments by casualty figures is measuring the wrong side of the transaction.
The question that measures the government is: at the end of it, does the state know who did this? For most of the six hundred, the honest answer is no — and for several hundred more it once thought it knew and was wrong.
That question is the one Chapter Four asks, and the answer it gets is worse than not knowing.
Twelve major attacks in about four years. Roughly six hundred dead.
Do not count the bodies. Count the convictions that are still standing — and there are very few. Men were held nineteen years for the Mumbai train bombings and acquitted in 2025.
And one set of these attacks hit mosques, shrines and a train to Pakistan. The men first arrested for those were the wrong men. That is the next chapter.
4The Wrong Men
On the eighth of September 2006, the eve of Shab-e-Barat, a series of explosions went off beside the Hamidia Mosque and the Bada Qabristan in Malegaon, a textile town in Maharashtra. Around thirty-seven people were killed and more than a hundred injured. Almost all of them were Muslim.
Within weeks the Maharashtra Anti-Terrorism Squad arrested nine men. All nine were Muslim.
4.1 — What was wrong with the case
The problems were not subtle and they were not discovered decades later. They were available at the time.
The Central Bureau of Investigation, which later took over the case, established two things about the men the ATS had named.
One of the accused was already in jail on the day of the bombing.
Another was seven hundred kilometres away, leading Friday prayers in a village, with witnesses. That is roughly the distance from Delhi to Bhopal.
These are not fine points of evidence requiring forensic expertise. They are the first two questions any investigator asks: where was he, and can anyone confirm it. The answers existed in 2006 and were available to anybody who asked.
The CBI found that the ATS had been given eyewitness accounts placing the man elsewhere and had chosen to disregard them.
It is worth being exact about where those two findings come from, because a hostile reader will assume they come from the men’s own lawyers.
Everything in the paragraph above comes from the state’s own investigators, not from the defence and not from journalism.
The Central Bureau of Investigation took the case over from the Maharashtra Anti-Terrorism Squad and recorded that the ATS had named a man who was in custody elsewhere on the day, and another whom eyewitnesses placed seven hundred kilometres away. The National Investigation Agency then filed a charge sheet in 2013 naming different people entirely.
So the account of what the first investigation got wrong is the account given by the second and third investigations, both of them agencies of the central government.
Part One set the rule that a charge built from the state’s own documents cannot be answered by disputing the documents. This chapter is the clearest instance of it in the series: to reject the finding, the government would have to disown two of its own agencies.
The nine men were held under the Maharashtra Control of Organised Crime Act, which makes bail extremely difficult and allows confessions made to police officers to be used in evidence — a provision that exists in no ordinary Indian criminal law, for the obvious reason.
Under ordinary Indian law, a confession made to a police officer cannot be used against the person who made it. The rule exists because confessions extracted in police custody are unreliable, and everybody involved in drafting the law knew it.
Special laws suspend that rule. Part Ten described TADA, used in Punjab, under which the conviction rate ran at one to two per cent — meaning that ninety-eight times in a hundred, the state arrested somebody it could not convict. MCOCA in Maharashtra works on the same principle.
When police confessions are admissible, the cheapest way to build a case stops being investigation.
4.2 — How long it took
The sequence, with the dates, because the dates are the charge.
| When | What happened |
|---|---|
| Sept–Nov 2006 | Nine Muslim men arrested by the Maharashtra ATS under MCOCA. |
| 2006–2010 | Held without trial. Bail refused. |
| Dec 2010 | Swami Aseemanand confesses before a magistrate that Hindu extremist groups carried out the attack. |
| Mar 2011 | Bail applications on the strength of that confession are rejected. |
| Nov 2011 | Bail finally granted, after the National Investigation Agency tells the court it will not oppose it. Five years in custody. |
| May 2013 | The NIA files a charge sheet naming a Hindu group. The Union Home Ministry publicly calls the original arrests “extremely unfortunate”. |
| Apr 2016 | The men are formally discharged — nearly ten years after arrest. One had died in the meantime. |
Two of the nine remained in custody after the 2011 bail order because they were also accused in the July 2006 train bombings — the case in which the Bombay High Court acquitted everybody in 2025.
4.3 — Who is charged with this, and who is not
Precision matters here more than anywhere else in this part, and this book is going to give the defence more than it usually gets.
The arrests were made by a state police unit. Maharashtra was governed by a Congress-led coalition, so the party is answerable for it, but the central government did not order those arrests and there is no evidence that it knew of them in advance. Alleged claims to the contrary have never been established and this book does not make them.
More than that: the central government’s own agencies are what unwound it. The CBI found the alibis. The NIA filed the charge sheet exonerating the nine. A Union minister said in public that innocent men had been jailed.
The defence says, and it is strong: this is a case of the system correcting itself. A state police unit made a catastrophic error. Two central agencies, under this government, investigated, established the truth and said so publicly. The National Investigation Agency, which did the correcting, was created by this government in December 2008. A government that builds the body that exposes its own state police units is not the villain of this chapter.
The reply, in three parts.
First, on the timing. The exculpatory facts — a man in jail, a man seven hundred kilometres away — did not require a new agency. They required somebody to read the file. Nothing was discovered in 2011 that could not have been discovered in 2006.
Second, on the bail refusal of March 2011. By then a magisterial confession pointing elsewhere existed and the men were still refused bail. That was a judicial decision, not a governmental one — but the state opposed the applications, and the state was answerable to this government’s law officers.
Third, and this is the one that does not go away: the Union Home Ministry called it “extremely unfortunate” in 2013. Nobody was disciplined. No officer faced proceedings. The men were not compensated by Maharashtra. An apology with no consequence attached is a press statement.
4.4 — Batla House
One case from these years is genuinely unresolved and this book is not going to pretend otherwise, because Chapter Four would be dishonest if it only contained cases that came out the same way.
On the nineteenth of September 2008, six days after the Delhi bombings, Delhi Police went to a flat in Batla House in Jamia Nagar. In the encounter that followed, two young men in the flat were killed, as was Inspector Mohan Chand Sharma. Two others were arrested and one escaped.
The police account is that the men opened fire and the officer was killed in the exchange. That account has been challenged since, on the grounds of the trajectory of the wounds, the absence of a video record, and the fact that no magisterial inquiry was ordered at the time.
The National Human Rights Commission examined it and found no violation. Courts have convicted two men in connection with the events at the flat. The challenges have not been sustained in any court.
Alleged in both directions, and this book records it as unresolved rather than picking the version that suits Chapter Four.
The temptation in a chapter about wrongful arrests is to add a case where the state killed rather than jailed. Part Six set the rule that a source is not promoted because its conclusion is convenient, and the same applies to a case.
What Batla House does establish is narrower and is not contested by anybody: that no magisterial inquiry was held at the time into a police operation in which three people died. That is a procedural failure whatever the underlying facts were, and it is the reason the argument has run for eighteen years with no way to end it.
4.5 — Hyderabad, and the money that was taken back
The same shape, in Andhra Pradesh, with one extra turn.
After the Mecca Masjid bombing of May 2007, in which nine worshippers died at Friday prayers and five more were shot dead by police in the crowd outside afterwards, a number of Muslim men were picked up and later released without charge.
The Andhra Pradesh government — Congress-run — then did something almost unique in this series. It paid them compensation for wrongful arrest.
And then the payment was challenged in court, and in September 2013 a division bench of the Andhra Pradesh High Court struck the compensation order down and directed the government to recover the money already paid. The order was suspended weeks later after the men themselves petitioned, and the state said it would appeal.
Everybody treats a wrongful arrest as a failure of the system. The word “failure” is doing enormous work there and almost nobody notices it.
A failure is a system not doing what it is built to do. So ask what this system is built to do, and how it measures itself.
A police force is assessed on cases detected and arrests made. A case is recorded as solved when somebody is charged. Conviction happens years later, in a different building, and is nobody’s performance indicator. Under a law that admits police confessions, the cheapest route to an arrest is a confession, and the cheapest route to a confession does not involve investigation at all.
By its own measures, the Malegaon 2006 investigation was a success. It produced nine arrests within weeks of a mass-casualty bombing, under a statute designed to make them stick. Every officer involved was doing what the system rewards.
That is worse than failure, and it is why Chapter Nine’s remedies are about measurement rather than about intent. Nobody has to be a bigot for this outcome to be produced reliably, which is exactly why it was produced reliably.
4.6 — What India does not have
There is one fact that sits underneath this whole chapter and it is usually left out because it sounds technical.
India has no law providing compensation for wrongful prosecution. Not a weak one. None.
A person acquitted or discharged after years in custody has no statutory entitlement to anything. He can bring a civil suit, which takes years he has already lost, or petition a High Court under its writ jurisdiction, which sometimes produces a payment as a matter of discretion. There is no scale, no presumption and no obligation on the arresting authority.
The nine men arrested for Malegaon in 2006 were held for five years before bail and discharged after nearly ten. One died before the discharge.
A Union minister said publicly that they were innocent. Their statutory entitlement on discharge was nothing. Not a rupee, not an apology on the record, not a reference to anybody for disciplinary action.
In Hyderabad, where a state government did pay compensation to men wrongly held, the payment was challenged and a High Court ordered it recovered before that order was suspended.
So the position is worse than an absence of a remedy. It is that a government which chooses to pay may be ordered to take the money back.
The Law Commission of India recommended a statutory scheme for compensation for wrongful prosecution in 2018. It has not been enacted. It appears again in Chapter Nine’s table of rules that were not made, and it is the one on that list that would have cost this government least and helped these men most.
4.7 — What is charged here
Established: that nine men were arrested in 2006 for a bombing they did not commit; that exculpatory evidence existed at the time and was disregarded; that they were held five years before bail and nearly ten before discharge; that a central agency of the same government later filed a charge sheet saying they were innocent; and that a Union minister said so publicly. Court-found and Commission-found in substance, and admitted by the government itself.
Not established, and not alleged here: that the central government directed or knew of the arrests. There is no evidence and this book does not assert it.
What is charged against the central government is narrower and survives every defence: that it presided for five years over a special-powers statute that made this outcome cheap; that it opposed bail after a magisterial confession pointing elsewhere existed; and that having admitted the men were innocent, it imposed no consequence on anybody and paid them nothing.
The last of those is the one Chapter Nine returns to, because the remedy is ordinary legislation and it still does not exist.
Nine men, arrested in 2006 for the Malegaon bombing. One was in jail that day. One was seven hundred kilometres away leading prayers.
They were held five years before bail and nearly ten before discharge. One died first. In 2013 the Union Home Ministry called it “extremely unfortunate”.
Nobody was disciplined. Nobody was compensated by the state that jailed them. And in Hyderabad, where a Congress state government did pay compensation to men wrongly held, a High Court ordered it recovered.
5Sixty Hours
Ten men left Karachi by sea, hijacked an Indian fishing trawler, killed its crew, and came ashore at Mumbai in an inflatable boat on the evening of the twenty-sixth of November 2008.
They split into groups and attacked a railway terminus, two luxury hotels, a hospital, a café and a Jewish centre. The last of them was killed on the morning of the twenty-ninth. A hundred and sixty-six people were killed, twenty-six of them foreign nationals, and more than three hundred injured. Nine attackers died. One, Ajmal Kasab, was captured alive.
This chapter does not narrate those sixty hours minute by minute. That has been done at book length by better-placed people and repeating it would add nothing. What this chapter does is set out what the state’s response reveals, because that is what a book about governments is for.
5.1 — How they arrived
The route matters, because every stage of it was a place the attack could have been stopped and was not.
The ten men left Karachi by boat. In the Arabian Sea they took an Indian fishing trawler, the Kuber, killed four of its five crew and forced the captain to sail them towards Mumbai. Near the coast they killed him too, transferred to an inflatable dinghy, and came ashore at Badhwar Park in Colaba — a fishing colony in the middle of the city — at around eight in the evening.
Local fishermen noticed them. One asked who they were and was told to mind his own business. He reported it. Nothing came of the report in time.
India’s coastline is about seventy-five hundred kilometres long. In November 2008 there was no single body responsible for watching it: the Navy, the Coast Guard, the marine police of each coastal state and the customs service each had a piece, and none had the whole.
Ten armed men crossed an international maritime boundary, hijacked a registered Indian vessel, sailed it for most of a day, and landed in the country’s largest port city without being intercepted.
The Kuber was found abandoned the next morning, with a body on board and a satellite phone left behind. That phone is how the investigation established where they had come from. The single most important piece of evidence in the case was left behind by accident.
None of that account was reconstructed afterwards from guesswork, and it is worth saying where it comes from before the chapter goes any further.
The sequence above is unusually well established for an event of this kind, and the reason is that one attacker was taken alive.
The route from Karachi, the hijacking of the Kuber, the landing point and the assignment of targets were all put in evidence at Kasab’s trial, tested by cross-examination, and upheld on appeal to the Supreme Court. The satellite phone left aboard the trawler, the interception of the handlers’ calls during the attack, and the recovery of the dinghy corroborate it independently.
Almost nothing else in this part rests on a foundation that solid — and Chapter Six, on what was known beforehand, rests on one of the weakest, because the relevant report has never been published.
5.2 — The response, in order
The attacks began at about half past nine in the evening. For the first stretch the police treated what was happening as a gang shooting.
It was past midnight — around three hours in — before the Chief Minister of Maharashtra asked the central government to send the National Security Guard.
There was no NSG unit anywhere near Mumbai. The commandos had to be flown from Manesar, near Delhi. When the NSG chief asked for an aircraft he was told the transport plane was in Chandigarh.
The commandos began operations at the hotels at around six in the morning. Roughly ten hours after the first shots.
Mumbai is India’s financial capital, a coastal city that had been bombed in 1993 and again in 2006. It had no rapid-response counter-terrorist unit of its own.
The police who did engage were badly outmatched, and not by a small margin.
An attack designed so that the attackers do not expect to survive, and are not trying to. The purpose is not to plant a bomb and escape; it is to keep shooting for as long as possible.
This matters operationally. Every Indian police response doctrine of the time assumed a hostage situation with people who wanted to negotiate and get out alive. Against attackers with no exit plan, containment and negotiation are not tactics. Someone has to go in.
The official inquiry that followed found that the Mumbai police were carrying weapons that were, in substance, of Second World War design, against attackers with modern assault rifles and grenades. Senior officers were found to have made significant errors of command. Three senior police officers were killed, including Hemant Karkare, the head of the Anti-Terrorism Squad — the man who, six weeks earlier, had made the arrests described in Chapter Four.
5.3 — The man who stopped one of them
Two of the attackers, after shooting into the crowd at the Chhatrapati Shivaji Terminus railway station and killing more people there than at any other single site that night, drove away in a hijacked car and were stopped at a police roadblock at Girgaum Chowpatty.
Tukaram Omble, an assistant sub-inspector, went at the car unarmed and took hold of the barrel of a rifle. He was shot at close range and killed. His colleagues took the man holding the rifle alive.
That man was Ajmal Kasab. He is the only attacker who was captured, and the entire evidentiary account of the planning of 26/11 — the Karachi departure, the training, the handlers on the telephone — rests on his interrogation and trial.
Everything India was later able to prove about who organised this attack, and everything it took to the rest of the world, exists because one police officer with no firearm decided to grab a rifle barrel instead of taking cover.
The equipment failure described in the next paragraph and the individual conduct described in this one are the same story told twice. The state supplied the men and not the means, and where the outcome was good it was because somebody made up the difference with his life.
Omble was awarded the Ashoka Chakra posthumously.
5.4 — What was done afterwards
The credit column here is real and this book gives it in full.
The Home Minister, Shivraj Patil, resigned on the thirtieth of November, taking what he called moral responsibility. He was replaced by P. Chidambaram. Within a month Parliament passed the National Investigation Agency Act, creating a national counter-terrorism investigating body for the first time. NSG hubs were established in four cities so that commandos would not have to be flown across the country again. Coastal security was reorganised. A national intelligence grid was begun.
The defence says: judge a government by what it does after a catastrophe, because that is the part it controls. Within five weeks of 26/11 this government had accepted a ministerial resignation, replaced the minister, created a national investigation agency, decentralised the commando force and begun rebuilding coastal security. Compare that with Part Fourteen’s Liberhan Commission, which took seventeen years to produce a document with no consequences.
It adds a harder point: the attack came from another country’s soil, planned and directed from there. No amount of Indian administrative reform prevents ten trained men arriving by sea from a neighbouring state that will not act against them. Blaming a government for being attacked confuses the crime with the response to it.
Where the defence wins: on the five weeks, substantially. The NIA is the most consequential institution created in these five years and Chapter Four shows it doing exactly what it was built for. And the point about the origin of the attack is correct.
Where it fails: everything in that list was available on the twenty-fifth of November. None of it required an attack to think of. The next chapter is about what was known before, and it is the reason this defence does not survive.
5.5 — The other nine, and the men who sent them
Nine of the ten attackers died in Mumbai. Everybody who planned the attack was in Pakistan, and none of them has ever been convicted anywhere for it.
Two people connected to the planning have been dealt with outside Pakistan. David Coleman Headley, an American who scouted the targets in Mumbai on several visits before the attack, was arrested in Chicago in 2009, pleaded guilty and was sentenced by a United States court in 2013 to thirty-five years. Under a plea agreement he was not extradited to India, though he later gave evidence to an Indian court by video link.
Tahawwur Hussain Rana, a Canadian citizen of Pakistani origin and a long-standing associate of Headley, was ordered extraditable by a court in California in 2023. After a series of appeals — the last of them rejected by the United States Supreme Court — he was flown to Delhi on the tenth of April 2025 and arrested on arrival by the National Investigation Agency. He faces ten charges including conspiracy and murder, denies all of them, and his trial is running.
From the attack to the arrival of one accused planner on Indian soil: sixteen years and four months.
He was not surrendered by the country the attack was launched from. He was extradited from the United States, after litigation in American courts, on the strength of a treaty signed in 1997.
In Pakistan, proceedings against those accused of organising the attack have run since 2009 without producing a conviction for it. The founder of the organisation India holds responsible was convicted there in 2020 — on terrorism financing charges, not for Mumbai.
Two things follow from that, and they cut in opposite directions.
The first is the point Chapter Five’s argument box already made: a government does not choose whether it is attacked from another state’s territory, and no amount of domestic reform reaches the people who planned this.
The second is that the agency which secured the 2025 extradition — the National Investigation Agency — is the institution created by the government of this part, in the five weeks after the attack. The single most substantial piece of accountability for 26/11 was obtained sixteen years later by a body UPA I built, under a government of a different party. Both halves of that sentence belong in the record.
5.6 — The trial
Ajmal Kasab was captured alive, tried, convicted, sentenced to death, and hanged on the twenty-first of November 2012. His trial produced the fullest account of the attack’s planning that exists in an Indian court record.
It also produced something this series should note, because it is unusual. He was given a defence lawyer at state expense, over public objection, and the trial ran through appeals to the Supreme Court. Part Ten described TADA courts in Punjab with conviction rates of one to two per cent and in-camera trials. This was the opposite: a full public trial for the most hated man in the country, and it held up on appeal.
That is a mark in the state’s favour and it is recorded as one.
So the defence carries this chapter and loses the next one, which is the honest division and is why they are two chapters rather than one.
Ten men. Sixty hours. A hundred and sixty-six dead.
Three hours before the state asked for commandos. Ten hours before they reached the hotels, because they had to be flown from near Delhi and the aircraft was in Chandigarh.
Afterwards: a minister resigned, an agency was created, commandos were based in four cities. All of that was possible on the twenty-fifth of November. The next chapter is about why it wasn’t done then.
6Seventeen Warnings
After the attacks, the Maharashtra government appointed a two-member committee under R. D. Pradhan, a former Union Home Secretary, with the former intelligence officer V. Balachandran. It reported within about four months.
Its central finding is the reason this chapter exists.
Commission-found. The High Level Enquiry Committee on the 26 November attacks found:
That relevant intelligence inputs were ignored or not acted upon seriously.
That there was total confusion in the processing of intelligence alerts at the level of the state government.
That inputs were forwarded to operational units mechanically, and that the machinery for making an overall assessment of them was inadequate.
That proper analysis of the available inputs would have revealed a strong indication of the attack.
That the police were ill-prepared and ill-equipped, and that senior officers made command errors.
The number attached to those alerts is the thing to hold on to.
The former head of state intelligence told the committee that there had been seventeen alerts since the seventh of August 2006 warning of the possibility of sea-borne attacks and of multiple simultaneous fidayeen assaults.
Seventeen alerts over twenty-seven months. Roughly one every seven or eight weeks, for more than two years, describing the shape of the attack that came.
He also told the committee that although he was the central intelligence authority in the state, central intelligence reports did not reach him directly, and he was therefore unaware of them.
Read those two paragraphs together. The warnings existed. The official whose job was to assess them did not receive them. And the reason he did not receive them was not that anybody withheld them — it was that no route existed.
Everybody called 26/11 an intelligence failure. The Prime Minister used the phrase. So did the opposition. So did most of the press.
But there were seventeen alerts. The intelligence did not fail. It was produced, and it was correct.
The assumption buried in the phrase is that intelligence means information. It does not. Information is raw. Intelligence is the act of deciding which information matters — of taking seventeen warnings spread over two years and saying: these three, together, mean something is coming.
That act is a job. It has to be somebody’s job, with a name on it. In November 2008 it was nobody’s. The alerts were forwarded mechanically, which is to say passed along by people whose duty was discharged by passing them along.
Part Thirteen found an addition nobody had made, where every component existed and no decision existed that anybody had to defend. This is that shape again, and here it killed a hundred and sixty-six people. Nobody ignored a warning. Everybody forwarded it. There was no desk at which forwarding stopped and deciding began.
Calling it a failure of intelligence puts the blame on the people who got it right.
6.1 — The Centre’s own share
The committee located the confusion at state level, and this book has already conceded that. But the arrangement the state was confused by was a central creation, and that has to go on the page.
After the Kargil conflict of 1999, a review committee found that Indian intelligence agencies did not share what they knew. A Multi-Agency Centre was created in 2001 to fix exactly that — a body where the intelligence services would pool material and produce a common assessment.
By 2008 it was not doing that job. It had no statutory basis, no power to compel any agency to give it anything, and no authority over what the states did with what it passed on.
The proof is what was built afterwards. In the months following 26/11 the Multi-Agency Centre was substantially strengthened, given round-the-clock operation and subsidiary centres in the states, and a national intelligence grid was begun.
Governments do not rebuild a working institution. The reforms of 2009 are an admission about 2008, and they are the most reliable evidence available because nobody had to be persuaded to make them.
The same logic applies in reverse to the Kargil review of 1999: its central recommendation was made, in name, in 2001 — and seven years later the thing it created could not do the thing it was created for.
So the honest division of blame is this. The state government could not process seventeen alerts. The central government had spent seven years not giving anybody the authority to make it possible. Neither of those is the other one’s excuse.
6.2 — The report the public was not allowed to read
The Pradhan Committee reported to the Maharashtra government — Congress-led — in 2009.
It was not published. The government tabled selected findings and an action-taken report in the state assembly and withheld the rest, citing security and the continuing trial of Kasab.
Kasab was hanged in November 2012. The report has still not been released in full.
Almost everything quoted in this chapter reached the public through leaks to newspapers and through what the committee’s own members later said, rather than through publication.
That is why the findings above are given as substance rather than as quotation, and why the seventeen-alert figure is attributed to what a witness told the committee rather than to a printed page anybody can check.
This book is telling you that its evidence here is second-hand, and that this is the state’s doing rather than the author’s. The primary document exists. It is ninety pages. It is seventeen years old. It is not public.
The reason given at the time — the ongoing trial — expired in 2012. No reason has been given since.
The defence says: a report of this kind necessarily describes exactly how intelligence is collected, routed and assessed, and how a security response is organised. Publishing it publishes a manual for the next attacker. Governments of every party withhold such documents and this is not a Congress practice.
That last sentence is true and this book concedes it. Part Six recorded the Henderson Brooks report on 1962 withheld across every government since; Part Fourteen recorded nine days of Home Ministry file notes from 1992 that no government of any party has released.
Where the defence fails: it is an argument for redaction, not for suppression. A ninety-page report can be published with operational detail removed, which is what other democracies do with equivalent inquiries. And the specific reason given in 2009 — the pending trial — has been spent for fourteen years without a replacement being offered.
Seventeen years is long enough for a security reason to become a habit, and habits do not have to be justified to anybody.
It was not an intelligence failure. There were seventeen alerts in twenty-seven months describing sea-borne, multiple, suicidal attacks. The intelligence was produced and it was right.
What did not exist was anybody whose job was to add them up. They were forwarded mechanically. The state’s own intelligence chief was not receiving them.
The official who found all this wrote a ninety-page report in 2009. A Congress state government did not publish it, and no government since has.
7The Deal
On the eighteenth of July 2005, in Washington, Manmohan Singh and George W. Bush announced that the United States would seek to resume civil nuclear cooperation with India.
To understand why that was extraordinary, you need one piece of background, and it is the only technical passage in this part.
Most countries have signed the Nuclear Non-Proliferation Treaty, which says: you may have civil nuclear power, but you may not have nuclear weapons. India refused to sign, on the ground that the treaty permits weapons to five countries and forbids them to everybody else.
India then tested a device in 1974 and again in 1998. In response, the countries that supply nuclear technology formed a group — the Nuclear Suppliers Group — which agreed not to sell nuclear fuel or reactors to any country outside the treaty.
So for thirty years India was locked out of the world market for nuclear fuel, while running a nuclear power programme and a weapons programme with domestic uranium.
What the 2005 statement proposed was that India would separate its civil reactors from its military ones, place the civil ones under international inspection, and in exchange be allowed to buy nuclear fuel and technology on the world market — without signing the treaty and without giving up its weapons.
No other country has ever been given that.
7.1 — What it took
| Date | Step |
|---|---|
| 18 July 2005 | Joint statement in Washington. |
| March 2006 | India’s separation plan: which reactors are civil, which are military. |
| December 2006 | The Hyde Act signed into US law, permitting cooperation with India. |
| August 2007 | The 123 Agreement between the two governments published. |
| 1 August 2008 | The IAEA Board adopts an India-specific safeguards agreement, unanimously. |
| 6 September 2008 | The Nuclear Suppliers Group grants India a waiver, after a first meeting in August ended without agreement. |
| 10 October 2008 | The agreement signed. |
Between the fifth and sixth rows falls the political crisis that occupies the rest of this chapter and all of the next.
7.2 — The objections, which were serious
Two very different sets of people opposed the deal and this book takes both seriously, because the government’s own supporting parties were among them.
The Left parties argued that the arrangement tied India into a strategic relationship with the United States, that the Hyde Act contained conditions about Indian foreign policy — particularly towards Iran — and that a domestic American statute would end up constraining Indian choices.
The Bharatiya Janata Party, which had itself begun the process of engagement with Washington in office, argued that the deal compromised India’s ability to test a nuclear weapon again, because fuel supply could be terminated if India tested.
The government’s answer was that the Hyde Act binds the United States and not India, and that nothing in the agreement removed India’s right to test — while conceding that testing would have consequences.
On the twelfth of September 2008, with the deal before the US Congress, President Bush stated that the fuel supply assurances given to India were political commitments rather than legally binding ones.
That was said by the deal’s principal author, in public, at the moment of ratification, and it went to precisely the point the Indian opposition had been making for three years.
A prosecution document should record when the other side’s objection turns out to have been correct in substance. This is such a case, and it does not depend on any Indian source.
The objections deserve to be put at their full strength rather than summarised, because the parties making them brought down a governing coalition over it and because one of them turned out to be right.
The Left’s case, at its strongest: this was never really about electricity. Nuclear power supplied about three per cent of Indian generation and was not going to supply much more in any plausible timeframe. What the deal actually bought was a strategic alignment with the United States, and it was bought at a moment when Washington’s regional priorities — Iran, Afghanistan, Pakistan — did not match India’s.
The Hyde Act was the evidence. It is a domestic American statute containing language about Indian policy towards Iran, and it is the law under which the American side would judge whether to keep supplying fuel. India would have built reactors dependent on imported fuel, and the supply of that fuel would rest on annual American determinations. That is not a commercial arrangement; it is a hostage. A country that has just made itself dependent has fewer choices, not more.
The BJP’s case, which pointed the other way: the deal quietly ended India’s ability to test again. Not in terms — nothing in the agreement forbids a test — but in consequence, because a test would terminate the cooperation and strand whatever had been built on imported fuel. A right you cannot exercise without ruining yourself is not a right.
The government’s answer: the Hyde Act binds the United States, not India; nothing in the agreement removes the right to test; and thirty years of exclusion had cost more than any of this.
How it actually turned out. On the first objection, the record is mixed — India did not follow American policy on Iran in the way the Left predicted, but the alignment they described did deepen through the following two decades under governments of both parties. On the second, the BJP’s point stands: India has not tested since 1998, and the constraint the deal created is one reason among several.
And on the specific question of whether the fuel assurances were binding, the objectors were vindicated within weeks — by the American President, in public. That is the next box.
7.3 — What it delivered
A book that puts an achievement in the credit column has to say what came of it, and here the honest answer is: less than was promised, and not for the reasons the opposition gave.
The deal opened the world market for nuclear fuel to India, and that part worked. Uranium supply for the existing civil reactors became reliable, which had been a genuine constraint.
What did not follow was the wave of new reactors. The reason is a second Indian statute, passed in 2010 under UPA II and therefore outside this part: the civil nuclear liability law, which made equipment suppliers liable in the event of an accident. Foreign suppliers declined to build on those terms. Two decades on, the number of American-built reactors operating in India is zero.
The deal was argued over for three years, brought down a governing coalition, and required a confidence vote to survive.
Nuclear power supplies roughly three per cent of India’s electricity, and did so before the deal as well.
The achievement was diplomatic rather than electrical. What India obtained in September 2008 was not power stations; it was the end of thirty years of being treated as an outlaw state, and formal acceptance that it could hold nuclear weapons without penalty. That is worth having. It is not what the country was told it was voting on.
7.4 — The rupture
On the eighth of July 2008, the Left parties withdrew support. The government lost its majority.
The Prime Minister could have shelved the deal. The safeguards agreement had not yet gone to the IAEA; the NSG had not met; nothing was irreversible. He chose instead to call a vote of confidence, on the tenth of July, and to go ahead.
This book has spent fourteen parts describing Congress governments doing what was necessary to stay in office. Here is one that put a policy ahead of survival, and said so, and staked the government on it.
On the substance: a genuine achievement. Thirty years of exclusion ended without signing the treaty and without giving up the weapons programme. India is the only country to have obtained this. It goes in the credit column without qualification.
On the politics: also to the government’s credit, and this book says so plainly. It risked office for a policy. That is rare enough in these nineteen parts to be worth naming.
The opposition’s central objection was partly vindicated by Bush’s own statement of 12 September 2008 about the fuel assurances, and this book records that too.
And then there is how the vote was won. That is the next chapter, and it is the reason this one cannot end here.
Which is to say the achievement is real, the objection was partly right, and the method of securing it is the charge. All three go on the page.
Three years, six international gates, and thirty years of exclusion from the world nuclear market ended — without signing the treaty and without giving up the bomb. No other country has been given this.
When the Left withdrew support in July 2008, the government could have shelved it. It called a confidence vote instead and staked office on a policy. In nineteen parts of this series, that is close to unique.
Then it won the vote. Chapter Eight is about how.
8The Vote, Again
On the twenty-second of July 2008, hours before the confidence vote, three Bharatiya Janata Party members of the Lok Sabha stood up in the House and emptied bundles of thousand-rupee notes onto a table.
They said the money — one crore rupees — had been given to them to abstain from the vote or to support the government, and named as the source an aide of the Samajwadi Party leader Amar Singh, whose party had just begun supporting the government after the Left withdrew.
The government won the vote, two hundred and seventy-five to two hundred and fifty-six. Twenty-one members cross-voted.
Readers of Part Fourteen will recognise every element of that paragraph, and this chapter is built on the recognition.
Everybody treats the cash-for-votes affair as an event of 2008. It has its own name, its own news cycle, its own court case.
Put it beside Part Fourteen, Chapter Eight, and the assumption breaks. July 1993: a Congress government without a majority faces a no-confidence motion, survives it, and is afterwards alleged to have paid members to vote with it. July 2008: a Congress government that has just lost its majority faces a confidence motion, survives it, and is alleged to have paid members to vote with it.
Same month of the year. Same trigger. Same party. Same allegation. Fifteen years apart.
Treating each as a scandal — a thing that happened, involving particular people who behaved badly — makes both unanswerable, because in both cases the particular people were never convicted of anything. Treating them as two instances of one mechanism makes the pattern the finding, and a pattern does not require anybody to be convicted.
The mechanism is this: an Indian government that loses its majority mid-term has one lawful route to survival, which is persuasion, and one unlawful route, which is purchase. The second is cheaper, faster and — until 2024 — was substantially protected by the Constitution. Part Fourteen established that protection. This chapter is what it looks like the second time.
8.1 — What was established, and what was not
The affair went through a parliamentary inquiry and a criminal investigation. Both are worth setting out, because they went in opposite directions.
A committee of the House under the Congress MP Kishore Chandra Deo inquired and reported in December 2008. It did not find the bribery allegation proved against Amar Singh. It did refer three men — an aide of Amar Singh’s, an aide of the BJP leader L. K. Advani who had organised the sting, and a BJP activist — to the Home Ministry for further action.
A parliamentary committee inquiring into an allegation against the government it supports has a structural problem, and it is worth naming rather than implying.
Such a committee’s membership reflects the strength of the parties in the House. The party with the most seats has the most members on it, and the committee that inquired here was chaired by a Congress MP.
Part Fourteen found the same shape twice: the Bofors joint committee of 1988 sat with a government majority and was boycotted by the opposition, and the securities-scam committee of 1992 produced regulatory reform and no punishment. Three committees, three decades, three governments, and in none of them did the House establish anything against the party running it.
That is not a claim that any of them was dishonest. It is a claim about what a body constituted that way is capable of finding, and the answer appears to be: nothing that costs its own majority anything.
The Delhi Police then did essentially nothing for two and a half years.
The money was produced on the floor of Parliament, on camera, in July 2008. There is video. The cash physically existed and was seized.
The first arrests were made in July 2011 — three years later — and only after the Supreme Court criticised the investigation as shoddy in a public interest case brought by somebody else.
Part Fourteen found that the Jain diaries sat in a CBI cupboard for five years until a journalist forced the government’s hand in the Supreme Court. This is the same sequence, with a shorter clock and better evidence. In this case the state was handed the physical money.
In August 2011 the Delhi Police filed a charge sheet naming Amar Singh, two of the three BJP members who had produced the cash, Advani’s former aide, Amar Singh’s secretary and a BJP activist. It alleged a conspiracy to deliver one crore rupees as illegal gratification.
Arrests followed in September 2011. Amar Singh went to Tihar jail and was bailed the following month on medical grounds.
8.2 — Where the money came from
This is the question the case turned on and never answered.
The Delhi Police told the court, in response to an application by one of the accused, that every effort had been made to trace the money and no evidence could be found. One crore rupees in cash appeared in Parliament and its origin was never established.
In March 2011, WikiLeaks published a United States diplomatic cable dated the twenty-first of July 2008 — the day before the vote. It recorded a conversation in which a senior Congress figure was said to have described a fund of fifty to sixty crore rupees set aside to secure support for the confidence motion.
Alleged. The cable is authentic as a cable — that is, it is a genuine record of what an American diplomat wrote down. What it records is a second-hand claim by an unnamed source about somebody else’s intentions.
That is a very long way from evidence that money changed hands. Part Six set the rule that a leaked document is graded on what it can support, not on how convenient its contents are, and it applies here.
What is established: that one crore rupees in cash reached three MPs, that this is on video, and that the state which seized the money never traced its origin.
What is not established: who provided it, or that anybody in the Congress leadership knew. This book does not assert either.
8.3 — How it ended
It ended the way Part Fourteen’s chapter ended, which is to say it did not really end at all.
The special court discharged six of the seven accused. The Delhi government later announced it would challenge the discharges in the High Court, on the ground that evidence had been ignored.
No one has been convicted. The three members who produced the cash were expelled by their party and two later returned to public life; one of them, Faggan Singh Kulaste, subsequently became a Union minister.
The defence says: a parliamentary committee found the allegation unproved. A criminal court discharged almost everybody. The money was produced by opposition MPs who had spent weeks arranging a sting operation with a television channel, and the charge sheet named the opposition leader’s own aide as the mastermind. On the evidence, the most that can be said is that a crore of rupees was moved around by people on both sides for the purpose of embarrassing each other.
That is a real argument and this book will not pretend otherwise. The sting was organised by people who wanted a scandal, and the man the police called the organiser was the opposition leader’s aide.
The reply: somebody paid for it. A sting requires the money to be genuinely offered, and one crore rupees was genuinely offered, in cash, by somebody. The state seized the notes and could not say where they came from. That is not a fact about the opposition’s motives. It is a fact about an investigation that began three years late and was described by the Supreme Court as shoddy.
8.4 — What changed in 2024, and what it means for both chapters
Part Fourteen ended its equivalent chapter with Sita Soren v. Union of India, decided on the fourth of March 2024, in which seven judges of the Supreme Court unanimously overruled the 1998 judgment that had given legislators immunity from prosecution for taking bribes to vote.
That judgment matters twice as much now.
| July 1993 · Part Fourteen | July 2008 · this chapter | |
|---|---|---|
| Trigger | No-confidence motion against a minority government | Confidence motion after allies withdrew |
| Result | Survived 265–251 | Survived 275–256 |
| Allegation | MPs paid to vote with the government | MPs paid to abstain or vote with the government |
| Investigation began | 1996, after the government fell | 2011, after the Supreme Court intervened |
| Money traced? | No | No — police told the court no evidence could be found |
| Outcome | Convicted 2000, acquitted 2002 | Six of seven discharged |
| Legal position at the time | Immune under Article 105(2) if they voted as paid | Same — the 1998 ruling still stood |
| Position now | Both would be prosecutable. Sita Soren, 4 March 2024, seven judges, unanimous. | |
The 2024 judgment is not retrospective and reopens nothing. What it establishes is that for the whole of the period covered by both chapters, the conduct alleged in both was constitutionally protected.
That is the finding of this chapter, and it is larger than either affair.
July 2008: one crore rupees in cash, produced on camera on the floor of the Lok Sabha, hours before a confidence vote the government won by nineteen.
The state seized the money and never found out where it came from. Investigation began three years late and only after the Supreme Court called it shoddy. Six of seven accused were discharged.
Now put it beside Part Fourteen, July 1993. Same month, same trigger, same party, same allegation, same non-result, fifteen years apart — and for both, the law of the land protected it until 2024.
8.5 — How it ended
Everything in this part happened before May 2009. In that month the government went to the country.
It was returned. The Congress won two hundred and six seats, up from a hundred and forty-five — its best result since 1991 — and formed a second government less dependent on its allies than the first.
Roughly six hundred people had been killed by bombs in the preceding four years. A hundred and sixty-six had died in Mumbai six months before polling. Nine men were in jail for a bombing the government’s own agencies were in the process of establishing they had not committed. Cash had been produced on the floor of Parliament ten months earlier.
The government’s majority went up by sixty-one seats.
Part Fourteen recorded the opposite outcome: a government that rescued the economy from default and was reduced to a hundred and forty seats. Two Congress governments, two records, and the electoral result runs in the opposite direction to the record in both cases.
A prosecution document has to reckon with that, and there are only three honest things to say about it.
The defence says: the electorate had all of this in front of it and returned the government with a substantially larger mandate. That is the only verdict in this book delivered by the people it is about. A writer who sets his own reading against theirs owes an explanation.
The reply has three parts, and only the third is strong.
The first, which this book does not use: that voters were misinformed. They were not. All of this was in the newspapers.
The second, which this book uses cautiously: elections are not referendums on a single record. In 2009 the alternative was a fragmented opposition, the employment guarantee was reaching tens of millions of households, and a farm loan waiver had been announced in 2008.
The third, which is the real one: the things in this part were not, in 2009, distributed the way votes are. Nine men in a Nashik jail are nine votes. The families in Malegaon and Hyderabad were, in each case, a small minority in a small number of constituencies. An electorate can deliver a fair verdict on how a government treated most people and no verdict at all on how it treated a few — and that is not a defect in the electorate, it is what majority rule is.
Which is why the remedies in Chapter Nine are all rules, and none of them is an election.
9What Would Have Prevented It
This chapter states the defence at full strength, says where it wins, and then sets out the rules that would have changed each outcome.
9.1 — The defence
One. The building record. Chapter Two. Two laws that changed what a citizen can demand of the state, a third recognising rights never recorded, and the first honest measurement of a minority’s condition. That is more constructive legislation than any government in this series since 1950.
Two. The attacks came from outside. Ten trained men arrived by sea from another country, planned and directed from there. Most of the bombings of 2005 to 2008 were organised across a border. A government does not get to choose whether it is attacked.
Three. It corrected its own state police. Chapter Four. The CBI found the alibis, the NIA filed the charge sheet exonerating nine innocent men, and a Union minister said so publicly. The agency that did it was created by this government five weeks after 26/11.
Four. It responded fast after November 2008. A ministerial resignation in four days, a new national investigating agency inside five weeks, commando hubs in four cities. Compare seventeen years for the Liberhan Commission in Part Fourteen.
Five. It risked office for a policy. Chapter Seven. When the Left walked out in July 2008, the deal could have been shelved at no cost. It was not.
Six. Almost nothing here is alleged against the central government directly. The wrongful arrests were made by a state police force. The intelligence-processing failure was found by the committee to sit at state level. The cash-for-votes prosecution named the opposition leader’s aide as organiser. A prosecution document that has to reach through two layers of government to find its defendant should say so.
Three of those six points come from bodies with no reason to help this government.
The Pradhan Committee located the intelligence-processing failure at the state level — and it was appointed by that same state government, which then declined to publish it, which is not what a government does with a report that exonerates it.
The Delhi Police charge sheet in the cash-for-votes case named the opposition leader’s aide as the organiser of the operation, and that charge sheet was filed only under Supreme Court pressure.
And the National Investigation Agency’s charge sheet exonerating the Malegaon nine was filed in 2013, under this government, contradicting two earlier investigations. An agency that overturns the state’s own earlier case is doing the opposite of what agencies are usually accused of.
9.2 — Where the defence loses
On three things, and they are connected.
The communal violence law. Promised in the Common Minimum Programme of May 2004. Drafted in 2005 and repeatedly afterwards. Never enacted, in ten years of office with a working majority for most of it. Part Eleven established what such a law would have to do — impose duties on officials, make command responsibility real, remove the requirement of political sanction to prosecute. Every year it was not passed is a year in which the machinery Part Eleven described stayed exactly as it was.
The consequence-free apology. Chapter Four. The government said in public that innocent men had been jailed for years. Nobody was disciplined. No compensation was paid by the state that jailed them, and where a state did pay, a High Court ordered it recovered. An admission with nothing attached is a press release.
The report. Chapter Six. Ninety pages, seventeen years old, still not published, the stated reason spent since 2012.
9.3 — The one that is hardest to answer
Of the three, the communal violence law is the one that does not go away, and it is worth setting out why rather than just listing it.
Part Eleven established what the machinery does. When a mob is at work, the police do not register the cases; untraced closure then follows automatically; acquittal follows from that. It is a coherent sequence rather than a failure, and every stage of it is lawful. Part Eleven’s finding was that this machinery survives every change of party, which is why voting differently is not a remedy for it.
A law was drafted in 2005 to break that sequence. It would have put a statutory duty on named officials to act, made command responsibility real so that a superintendent could be prosecuted for what his force did not do, and removed the requirement of political sanction before an officer could be tried.
It was drafted, redrafted and dropped repeatedly across ten years in office.
Between the promise of May 2004 and the end of that government in 2014 there were ten years and two comfortable majorities.
In that decade the same governments passed a transparency law, an employment guarantee, a forest rights law, a food security law and a land acquisition law — all of them contested, all of them opposed by organised interests, all of them enacted.
The one that would have imposed duties on police officers is the one that was never brought to a vote. That is not a failure of legislative capacity. This government had legislative capacity and used it repeatedly.
There is a real answer to that and it deserves stating properly, because it is the best thing the defence has on this point.
The defence says: the drafts were genuinely bad. The 2011 version defined the offences in terms of a “group” that was framed so as to apply asymmetrically, and was opposed on that basis by state governments of several parties and by legal commentators who were not hostile to the object. Law and order is a state subject, and a central statute imposing duties on state police invites a constitutional challenge it might not survive. A government that does not pass a defective bill has not failed; it has declined to pass a defective bill.
The reply: every one of those objections is an argument about drafting, and drafting is what ten years is for. The asymmetry point was made in 2011 and could have been fixed by 2012. And the federal objection is answerable, because Parliament has legislated repeatedly on criminal procedure, which is on the concurrent list.
What the defence keeps: that no other government has passed such a law either, before or since. Part Eleven said so and this part repeats it. The charge here is a broken promise, not a unique failing.
9.4 — The rules that were not made
| The failure | The rule that would have caught it | Status |
|---|---|---|
| Seventeen alerts, forwarded and never assessed | A named officer, at state and central level, statutorily responsible for aggregate threat assessment, reporting on a fixed cycle | Partly built after 2008 by executive action; never legislated |
| Nine men held five years on a case with visible holes | Statutory outer limit on pre-trial custody under special-powers laws, with automatic bail on expiry | Never enacted |
| Police confessions admissible under MCOCA | Repeal of the confession provisions in state organised-crime statutes | Never attempted |
| Innocence established, nothing paid | A statutory right to compensation on discharge or acquittal after prolonged custody, payable by the arresting authority | Recommended by the Law Commission; never enacted |
| Cash produced in Parliament, origin never traced | Automatic reference of any bribery allegation arising on the floor to an independent agency within a fixed period | Never enacted |
| Communal violence, no duty on officials | The bill this government drafted in 2005 | Drafted four times. Never passed. |
Every one of these is ordinary legislation. None requires a constitutional amendment. Parts Nine, Eleven, Thirteen and Fourteen each ended with a comparable table and Part Nineteen consolidates them all.
9.5 — The question turned on this book
This series assumes that a government’s failures and the state’s failures are the same thing. Every part has charged a government with what happened on its watch.
Part Fifteen is where that assumption comes under real pressure. Look at what this part actually charges. A state police force arrested the wrong men. A state intelligence system did not route its alerts. A state government withheld a report. A criminal justice system took nineteen years to acquit twelve men. A special-powers statute passed by a state legislature made confessions admissible.
None of those machines was built by this government, and none of them has been dismantled by any government since. They were built over decades, mostly by earlier Congress governments, which is how this series can charge the party as one continuous actor — and they have outlasted every change of party at the Centre, which is how the defence escapes.
So which is it? The honest answer is that this book cannot separate the two, and neither can anybody else. A government inherits a state. It is answerable for what it does with it and not for its existence. Where the line falls between those is a judgement, not a finding, and every reader will draw it somewhere different.
What this book can do is refuse to hide the problem. The charges in Chapters Four and Six are charges against Indian policing, and the specific charge against this government is that it had five years, a written mandate and a majority, and did not change it. That is smaller than the chapter titles suggest, and it is what the evidence supports.
Leaving that problem unresolved in the middle of the part is deliberate. Resolving it quietly, in this book’s own favour, would be the easier thing and the dishonest one.
The defence is the strongest in this series. Two transformative laws, a genuine foreign-policy achievement, a fast institutional response after 26/11, and central agencies that corrected a state police force’s worst error.
It loses on three: a law against communal violence promised in 2004 and never passed in ten years; an admission that innocent men were jailed, with no consequence for anybody; and a report still unpublished seventeen years on.
And the deepest problem is this book’s, not the government’s: most of what went wrong here was done by machines nobody in office built and nobody since has fixed.
10An Honest List of What We Do Not Know
Every part of this series ends here, and this one has more to put in the list than most, because several of its central cases are still moving.
10.1 — Not established
Who carried out most of the bombings of 2005 to 2008. For the great majority of roughly six hundred deaths there is no conviction standing today. In several cases convictions were obtained and reversed. This book does not assert who was responsible for any attack where a court has not found it.
Whether the central government knew of the Malegaon arrests. Not established, not alleged. Chapter Four says so twice.
Where the cash in Parliament came from. The state seized it and could not trace it. The WikiLeaks cable is a genuine record of a second-hand claim and nothing more.
Why Sonia Gandhi declined the office in May 2004. Principle, calculation, or the constitutional challenge that was being threatened. No account settles it and this book does not try.
What was in the Aseemanand confession’s retraction. He confessed before a magistrate in December 2010 and retracted later. Courts have since acquitted in the cases built on it. Which version is true is not established, and the acquittals mean that as a matter of law the answer is now: neither.
10.2 — Where this part may be wrong
The casualty figures in Table 3. They vary between sources by ten to twenty per cent for most of these attacks. There is no single authoritative national register of deaths in terrorist incidents that reconciles them.
The final column of Table 3 is a moving object. Several of those cases are under appeal. The 2025 acquittals in the Mumbai train case are being challenged. A reader in five years may find that column substantially wrong, and this book would rather say so now than be caught by it.
The seventeen-alert figure. It comes from what a witness told the Pradhan Committee, reported second-hand, because the report is not public. Chapter Six says this in the text rather than in a footnote. If the report is ever released and the number is different, this part is wrong on its most-quoted fact.
The placement of Chapter Two. As in Part Fourteen: putting the achievement early is a choice, and a hostile reader may say the credit is banked so the prosecution can proceed with clean hands. That reading is available.
10.3 — The count nobody keeps
There is a second missing thing in this part and it is not a document. It is an addition.
India has no single authoritative register that records, for each terrorist attack: how many died, who was arrested, what happened to each case, and whether anybody is serving a sentence today. The figures in Table 3 were assembled from court records and press reports and vary between sources by ten to twenty per cent.
Every component of that register exists. First information reports are filed. Post-mortems are done. Charge sheets are numbered. Judgments are published. Prisons keep rolls. Nothing is secret and nothing is missing. The addition has simply never been done.
Part Thirteen found exactly this shape about displacement — every number existing somewhere, no total anywhere, and no decision that anybody has to defend. It named that as a distinct form: an unmade addition, where concealment is not required because the absence of a sum is nobody’s act.
A published national register of terrorist incidents and their judicial outcomes: deaths, arrests, charge sheets, convictions, acquittals, and convictions surviving appeal.
It requires no new information. It requires somebody to be made responsible for adding up information the state already holds, and to publish the result annually.
Part Five ended by naming the one calculation that would settle freight equalisation, which nobody has done. Part Fourteen named a second. This is the third, and like the others it is arithmetic rather than a secret.
Until it exists, the sentence “most of the six hundred people killed between 2005 and 2008 have had no justice” is something this book believes on the evidence it could gather, and cannot prove.
10.4 — The document that would settle it
The report of the High Level Enquiry Committee on the 26 November attacks. Two members. Around ninety pages. Delivered to the Government of Maharashtra in 2009.
It is the only systematic official account of what was known before the attack, who was told, and where the assessment broke down. Its findings are in this part only because parts of it leaked and because its authors have spoken since.
The reason given for withholding it in 2009 was the pending trial of the surviving attacker. He was hanged in November 2012. No reason has been offered in the fourteen years since.
Part Six named the Henderson Brooks report, withheld since 1963. Part Fourteen named nine days of Home Ministry file notes from 1992. This is the third, and it is the youngest and the shortest of the three.
One closing observation, which is uncomfortable and belongs here.
The government that withheld that report is the same government that passed the Right to Information Act. Chapter Two called that Act the largest single transfer of power to ordinary people in this series, and meant it.
Both things are true. A government can build the best transparency instrument in the country’s history and then decline to apply it to the document that would embarrass it most, and there is no contradiction in that at all — because the Act contains an exemption for national security, and the exemption is invoked by the same officials the Act was meant to compel.
Part Thirteen found that a right is worth the independence of whoever enforces it. Here the same government supplied both the right and the enforcer, and the enforcer lost.
For most of the six hundred people killed between 2005 and 2008, nobody is serving a sentence and this book cannot tell you who did it.
What is missing is one document: ninety pages, written in 2009, about what was known before 26/11. Not a cabinet paper. Not an operational file. An inquiry report into a public catastrophe.
The reason given for withholding it expired in 2012. And it is being withheld by a government that passed the Right to Information Act.
Sources & further reading — Part 15
The State Ledger
Part Five committed every subsequent part to carrying this table: which states were affected by the material in this part, and how. It accumulates across the series and consolidates in Part Nineteen.
| State | What happened, 2004–2009 | Standing |
|---|---|---|
| Maharashtra | Malegaon 2006 and the nine wrongful arrests; the July 2006 train bombings; Malegaon 2008; 26/11; the Pradhan Committee report withheld. Congress-led state government throughout. | The central case of this part |
| Andhra Pradesh | Mecca Masjid, May 2007; men wrongly held then released; compensation paid by the Congress state government and then ordered recovered by the High Court. | Charge, with a partial credit |
| Rajasthan | Ajmer dargah, October 2007. Jaipur, May 2008 — death sentences in 2019, all acquitted on appeal in 2023. | Charge |
| Haryana | Samjhauta Express, February 2007. Sixty-eight killed, mostly Pakistani nationals. All accused acquitted 2019. | Charge |
| Delhi | Blasts October 2005 and September 2008. The cash-for-votes investigation, begun three years late. | Charge |
| Uttar Pradesh | Varanasi, March 2006. Convictions obtained in 2022 — sixteen years. | Charge |
| Gujarat · Karnataka | Ahmedabad and Bangalore, July 2008. Ahmedabad produced 38 death sentences in 2022, under appeal. | Charge |
| Every state | The Right to Information Act 2005; the rural employment guarantee from 2006; the Forest Rights Act 2006 (see Part Thirteen). | The largest credit item in the ledger |
Timeline
| Date | Event |
|---|---|
| 13 May 2004 | Congress wins 145 seats. UPA formed with Left support from outside. |
| 18 May 2004 | Sonia Gandhi declines the office. |
| 22 May 2004 | Manmohan Singh sworn in. |
| Sep 2004 | POTA repealed, as promised in the Common Minimum Programme. |
| Mar 2005 | Sachar Committee appointed. |
| 15 Jun 2005 | Right to Information Act passed. In force 12 October. |
| 18 Jul 2005 | Bush–Singh joint statement on civil nuclear cooperation. |
| Aug–Sep 2005 | National Rural Employment Guarantee Act passed. |
| 29 Oct 2005 | Delhi market bombings before Diwali. |
| 16 Jan 2006 | ₹717 crore package announced for the victims of November 1984 (Part Eleven). |
| 7 Mar 2006 | Varanasi bombings. |
| 11 Jul 2006 | Seven Mumbai suburban trains bombed. ~189 killed. |
| 7 Aug 2006 | First of the seventeen alerts later described to the Pradhan Committee. |
| 8 Sep 2006 | Malegaon. Nine Muslim men arrested within weeks. |
| Nov–Dec 2006 | Sachar report published. Forest Rights Act passed. Hyde Act signed in the US. |
| 18 Feb 2007 | Samjhauta Express bombed. 68 killed. |
| 18 May 2007 | Mecca Masjid, Hyderabad. |
| Aug 2007 | 123 Agreement published. |
| 11 Oct 2007 | Ajmer dargah bombing. |
| 13 May 2008 | Jaipur bombings. |
| 8 Jul 2008 | Left parties withdraw support. |
| 22 Jul 2008 | Cash produced in the Lok Sabha. Government survives 275–256. |
| 25–26 Jul 2008 | Bangalore and Ahmedabad bombings. |
| 1 Aug 2008 | IAEA adopts the India-specific safeguards agreement. |
| 6 Sep 2008 | Nuclear Suppliers Group grants India a waiver. |
| 12 Sep 2008 | Bush states the fuel assurances are political, not legally binding. |
| 13 Sep 2008 | Delhi bombings. |
| 29 Sep 2008 | Malegaon again. The ATS under Karkare arrests members of a Hindu organisation. |
| 10 Oct 2008 | Nuclear agreement signed. |
| 26–29 Nov 2008 | Mumbai. 166 killed. Karkare among the dead. |
| 30 Nov 2008 | Shivraj Patil resigns; P. Chidambaram becomes Home Minister. |
| Dec 2008 | National Investigation Agency Act passed. Pradhan Committee appointed 30 December. |
| Dec 2008 | UAPA amended, restoring most of what POTA had contained — without the confession clause and without an expiry date. |
| 2009 | Pradhan Committee reports. Not published. UPA returns to office with 206 seats. |
| Dec 2010 | Aseemanand’s magisterial confession. |
| Jul–Sep 2011 | Cash-for-votes arrests, after the Supreme Court criticises the investigation. |
| Nov 2011 | The Malegaon nine granted bail. Five years in custody. |
| 21 Nov 2012 | Kasab hanged. The stated reason for withholding the Pradhan report expires. |
| May 2013 | NIA charge sheet names a Hindu group for Malegaon 2006. Union minister calls the arrests “extremely unfortunate”. |
| 2013 | David Coleman Headley sentenced to 35 years by a US court. |
| Apr 2016 | The Malegaon nine formally discharged, nearly ten years on. |
| 2017–2019 | Acquittals in the Ajmer, Mecca Masjid and Samjhauta cases. |
| 2023 / 2025 | Jaipur convictions overturned; the Mumbai train accused acquitted after ~19 years; Malegaon 2008 accused acquitted. |
| 4 Mar 2024 | Sita Soren: seven judges end legislators’ immunity for bribery. Chapter Eight. |
| 10 Apr 2025 | Tahawwur Rana extradited from the United States and arrested by the NIA — sixteen years and four months after the attack. |
Words Used in This Part
| Approver | An accused person who admits the offence and gives evidence against the others, usually for a pardon. |
| Article 105(2) | The clause protecting MPs from court action for anything said or any vote given in Parliament. Read narrowly since Sita Soren (2024). |
| Charge sheet | The document by which police formally accuse named people of an offence and take the case to court. |
| Common Minimum Programme | The written list of commitments agreed between coalition partners in 2004. Table 1 measures the government against it. |
| Confidence motion | A vote on whether the government still commands a majority. Lost, it must resign. |
| Discharge | A court’s decision that there is not enough material even to put somebody on trial. Stronger than an acquittal, because the trial never starts. |
| Fidayeen | An attack in which the attackers do not intend to survive. |
| IAEA | The International Atomic Energy Agency, which inspects civil nuclear facilities. |
| MCOCA | Maharashtra’s organised-crime statute. Makes bail hard and admits confessions made to police officers. |
| Magisterial confession | A confession made before a magistrate, who must first warn the maker and be satisfied it is voluntary. Admissible where a police confession is not. |
| NIA | The National Investigation Agency, created by Act of Parliament in December 2008. |
| NSG (commandos) | The National Security Guard, India’s counter-terrorist force. Based at Manesar in 2008; hubs added in four cities afterwards. |
| Nuclear Suppliers Group | The countries that sell nuclear technology, which agreed not to supply states outside the non-proliferation treaty. Granted India a waiver in September 2008. |
| Right to Information | The 2005 Act giving any citizen the right to demand information from a public authority, with a personal penalty on officials who refuse. |
| Support from outside | Voting with a government without joining it. Influence without responsibility, withdrawable at any time. |
| Terms of reference | The written instructions creating an inquiry and limiting what it may examine. See Part Fourteen, Chapter Five. |
A Word on Sources
Everything dated in this part was checked against a source rather than recalled, and where sources disagree the disagreement is on the page.
The court record does most of the work here. Convictions, acquittals, discharges and bail orders in the Malegaon, Mecca Masjid, Ajmer, Samjhauta, Jaipur, Mumbai train and cash-for-votes cases are all matters of record. So is the National Investigation Agency’s 2013 charge sheet, and so is the Supreme Court’s criticism of the Delhi Police investigation. The nuclear deal’s stages are documented on both governments’ records.
The weakest evidence in this part is in Chapter Six, and the chapter says so where it stands rather than here. The Pradhan Committee’s findings are given as substance rather than quotation because the report has never been published; they reach the public through leaks and through what its authors have said since. The seventeen-alert figure has that status. If the report is released and contradicts this part, Chapter Ten has already said which sentence to strike.
Three things are graded Alleged and nothing rests on them: the WikiLeaks cable’s account of a fund set aside before the confidence vote, any claim that the central government knew of the Malegaon arrests, and the attribution of any bombing where no conviction now stands.
Casualty figures for the attacks in Table 3 vary between sources and are given as commonly cited, with the variation flagged in Chapter Ten rather than smoothed away.
A note from the author
I have found this the hardest part to keep honest, and the difficulty is the opposite of the one I expected.
I came to these five years with a settled view, formed the way most people’s views of the UPA were formed, which is backwards from what came after. Part Sixteen is the scandals part, and everybody knows what UPA II looked like by 2014. It is very easy to read the first term as a rehearsal for the second, and I had to stop myself doing it three times while drafting.
The place where that would have distorted the writing is Chapter Two. My first outline had the Right to Information Act and the employment guarantee as a short section in Chapter Nine, among the defences. That is where a prosecutor puts things he has to mention. Moving them to Chapter Two, at length, changed what the rest of the part could honestly say — and it should have, because on the evidence they are the largest constructive act by any government in this series since the 1950s.
The second place is Chapter Four, and there my problem was the reverse. That chapter is the strongest material I have, and I wanted it to be a charge against the central government. It is not. It is a charge against a state police force, a state statute and a system of measurement, and the central government’s own agencies are what exposed it. I have written it that way and it is weaker as advocacy and stronger as evidence.
Part Ten declared that I am Punjabi and had written on that material before. Nothing here touches that. What I will declare instead is that Chapter Four concerns men jailed on the basis of who they were rather than what they did, and that this is the kind of material where I am least likely to notice myself pressing too hard. The argument box in 4.3 gives the defence more room than any other in this part, deliberately, for that reason.
One last thing. Chapter Nine’s final box says this book cannot separate a government’s failures from the state’s. I do not have a solution to that, and I would rather leave the problem visible in the middle of the part than resolve it quietly in my own favour.
Lovepreet Singh
Part Fifteen of The Congress Record
The Living Archive · misterlove.in
Part Sixteen takes up UPA II, 2009–2014: 2G, Coalgate, the Commonwealth Games, Adarsh, and what the courts actually found afterwards. The comparison with the government of 2014 onwards is Part Eighteen, on the same ruler, and the second series auditing that government on this scale is still promised.