Skip to content Skip to footer

How McChrystal and Petraeus Built an Indiscriminate “Killing Machine“

Gen. Stanley McChrystal describes his strategy going forward in Afghanistan after President Barack Obama announced he will increase the US troop presence by 30,000 beginning in early 2010. (Photo: U.S. Army Sgt. David E. Alvarado / isafmedia)

Even if the rest of the US military effort in Afghanistan has been largely written off by the news media as a failure, the campaign of targeted raids against insurgents by commandoes under the Joint Special Operations Command (JSOC) has gotten a reputation for devastating effectiveness. As an Associated Press story in early September 2010 put it, the “mystique of elite, highly trained commandos swooping

down on an unsuspecting Taliban leader in the dead of night plays well back home….”

Central to this larger-than-life image of the Special Ops night raids in Afghanistan is the assumption that their targeting has been highly accurate.

John Nagl, who was on Gen. David Petraeus's staff in Iraq and now runs the pro-military think tank Center for New American Security, suggested in a PBS “Frontline” documentary on the targeted raids last January that the US military had gotten “so good at using electronic means of identifying, tracking and finding” insurgents that it had created an “industrial strength counterterrorism killing machine.”

The accumulated evidence that has now seeped through the cloak of secrecy surrounding Special Operations Forces (SOF) “kill/capture raids” tells a very different story, however. Although the raids have undoubtedly killed a large number of Taliban commanders and fighters, it is now clear that they also killed and incarcerated thousands of innocent civilians. The failure to discriminate between combatants and civilians flows directly from a targeting methodology that is incapable of such discrimination.

Creating Intelligence Without Human Vetting

The system of targeted raids now practiced in Afghanistan was first introduced by Gen. Stanley A. McChrystal in Iraq. In his book “The War Within,” Bob Woodward described how the JSOC under McChrystal had adopted “some of the most highly classified techniques” in the US government system of classification in Iraq in 2006 and 2007 to target and kill al-Qaeda and Shi'a militia fighters. Woodward contributed to the myth of all-knowing JSOC commanders by suggested that those techniques may have been the biggest factor in reducing the violence in Iraq. It is well documented, however, that Sunni and Shi'a insurgencies ended or tapered off because of broader political factors, not because of JSOC's killing operations.

In fact, McChrystal's operation relied on far more mundane technologies than Woodward's sensational language suggested. In a new book, “Task Force Black,” by Mark Urban, the diplomatic editor at BBC's “Newsnight,” reveals that McChrystal's command gathered intelligence on al-Qaeda and Mahdi Army personnel from three well-known technologies: 24-hour surveillance by drone aircraft, monitoring of mobile phone traffic and pinpointing the physical location of the phones from their signals.

The key to JSOC definition of a given insurgent “network” was the decision to maintain long-term aerial surveillance of a particular location. McChrystal's intelligence chief Col. Michael Flynn liked to call surveillance by drone aircraft “The Unblinking Eye” – an image suggesting a godlike power of observation. The implication of the new intelligence methodology developed by McChrystal and Flynn was that anyone who visited a location under surveillance or who communicated with a mobile phone associated with that location could be considered to be part of the insurgent network.

To convert the raw data obtained from drone surveillance and tracking mobile phone calls into intelligence on who to target, Flynn turned to a tool called “social network analysis.” In the wake of the 9/11 attacks, that technique had became almost overnight the favored tool for analyzing terrorist and insurgent networks. It provided a framework for construction of models of networks by measuring the number of direct interactions between individuals or “nodes.” With a quantitative tool called “link analysis” and accompanying software, intelligence analysts could see the raw data from drone surveillance and links among telephones transformed into a “map” of the insurgent “network” in each locality.

Traditional intelligence analysis of an insurgent network would have involved verifying the identities of those individuals who had visited the location or communicated with others associated with the network to assess the nature of the relationship. From the beginning of the new McChrystal-Flynn system, however, the emphasis was on speed of collection rather than on such careful analysis of the data. Urban recalls that McChrystal and Flynn introduced the concept of “F3EA” – “find, fix, finish, exploit, analyze.” That meant that they aimed at obtaining new data from each raid that could be used to add to the target list for future raids – often within hours of the previous one.

Scaling Up Raids in Afghanistan

In 2009, the whole system began to shift from Iraq to Afghanistan. McChrystal was chosen as the new commander of all US and NATO forces in Afghanistan and took Flynn with him. As the US began its drawdown in Iraq, thousands of SOF under JSOC command and most of the drones used in Iraq followed. Meanwhile, the focus of targeting of night raids in Afghanistan shifted from “high value targets” – high-level and mid-level al-Qaeda and Taliban officials – to anyone who was contributing to the Taliban war effort, whether in a military or civilian capacity.

These geographical and targeting shifts resulted in an exponential increase in the level of targeted raids in Afghanistan. In May 2009, before McChrystal arrived, US SOF were carrying out 20 raids per month. But by November, McChrystal had stepped up the pace to 90 per month, and by the following spring, he had increased the number again to nearly 250 a month – a 12.5-fold increase in one year.

Finally, during the transition from McChrystal to Gen. David Petraeus in the summer of 2010, the number increased to nearly 600 raids a month. In just two years, the monthly total of night raids had been expanded by a factor of 30. But in April 2011, a US military source told researchers for the Open Society Foundations and The Liaison Center that as many as 40 raids were being carried out every night – a rate of more than 1,000 raids per month.

Scaling up the system of targeted raids by orders of magnitude would have had far-reaching implications for the accuracy of the targeting regardless of the location. But it was especially dangerous to transfer it to Afghanistan. Michael Semple, former deputy to the European Union (EU) Special Representative to Afghanistan, who has had contacts with many local Taliban commanders over the years, told me that most Afghans in the Pashtun south and east “have a few Taliban commander numbers saved to their mobile phone contacts” as a “survival mechanism.”

That fact means that US intelligence analysts working on targeting for the SOF raids must be able to “distinguish pragmatic, innocent contacts from active involvement with the Taliban,” Semple told me. Unfortunately, the methods used by US intelligence to compile the list of supposed network members is not aimed at making such distinctions at all.

Targeting Phone Numbers, Not People

In October 2009, when JSOC was carrying roughly 90 raids per month, the target list for SOF night raids, called the Joint Prioritized Effects List (JPEL), included 2,058 names, according to one of the Afghan war logs documents released by WikiLeaks. A large proportion of the targets on the list were not identifiable individuals at all, but mobile phone numbers. “When you are relying on cell phones for intelligence, you don't get the names of those targeted,” observes Matthew Hoh, who was briefed on the target list as the senior US civilian official in Zabel province before he quit in protest against US policy in September 2009.

The easiest way to scale up the JPEL to support 900 raids a month was to get more mobile phone call records linked to numbers already associated with insurgent networks. One obvious source is the population of roughly 3,300 suspected insurgents being held in the Afghan prison system, who are allowed to use mobile phones freely in their cells. Semple told this writer that access to cell phones by the Afghan prison population is the result of the connivance of the prison administration, in return for bribes from the prisoners. “I presume there are occasional searches to keep the price up,” Semple said.

But the population of the security wing at Pol-e-Charkhi, the main Afghan national prison, is probably the largest concentration of insurgent-related cell phone users in Afghanistan, according to Semple, and therefore intelligence agencies would naturally seek to exploit the phone contacts among prisoners and friends and families with varying degrees of ties to the insurgents.

Another knowledgeable source in Kabul confirmed to me that Afghan corrections officers in Pol-e-Charkhi as well as in five regional prisons carry out “shakedowns” in the cells of security prisoners every six months or so to confiscate “contraband” – meaning primarily their mobile phones. These “shakedowns” are done at the prompting of contract corrections advisers and trainers funded by the US government, the source told me, and the confiscated phones are turned over to US intelligence.

The phone numbers and call histories from those phones go into the database which is used to “map the networks.” But the link analysis methodology employed by intelligence analysis is incapable of qualitative distinctions among relationships depicted on their maps of links among “nodes.” It operates exclusively on quantitative data – in this case, the number of phone calls to or visits made to an existing JPEL target or to other numbers in touch with that target. The inevitable result is that more numbers of phones held by civilian noncombatants show up on the charts of insurgent networks. If the phone records show multiple links to numbers already on the “kill/capture” list, the individual is likely to be added to the list.

Kandahar and other provinces are rife with stories of people who were targeted by night raids because their cell phone number had been found in an insurgent's cell phone. One Afghan told me about a friend whose brother had been seized and detained in a night raid and had been told by his interrogator that it was because of his cell phone calls to a known insurgent.

The US military's obvious lack of concern about targeting noncombatants is due in large part to the assumption that civilians have knowledge of insurgents that can be usefully exploited if they are brought in for interrogation. That was the motive for US and Canadian troops to sweep large numbers of military age males in their large-scale operations in 2004 and 2005. “The detainees are detained for a reason,” said Canadian Brig. Gen. Jim Ferron, the intelligence chief for NATO's International Security Assistance Force (ISAF) command in Afghanistan in May 2007. “They have information we need.”

The intelligence value of detaining large numbers of civilians with presumed knowledge of insurgents is now a motive for the intelligence analysts compiling the “kill/capture list” to catch civilians in the net. A new report on night raids by the Open Societies Foundation and The Liaison Center quotes a military officer telling the author privately, “If you can't get the guy you want, you get the guy who knows him.” And even when they are not the targets of the raid, civilians are deliberately swept up in order to interrogate them for several days before releasing them.

“A Very Precisely Targeted Operation”

Some raids, however, are deliberately aimed at killing the target, and cell phone data are also used to determine who is to be killed, as is dramatically illustrated by the killing of Zabet Amanullah by an SOF unit in Takhar province on September 2, 2010. The incident itself and the US military response to criticism of the killing as an obvious mistake highlights the reality that the intelligence analysts mapping the insurgent “network” routinely fail to make any effort to distinguish between an insurgent and someone who is in tangential contact with an insurgent.

Amanullah was a former Taliban commander who had quit the organization in 2001 and had become a human rights activist. He worked with former EU official Semple from 2003 to 2007, documenting pre-2001 human rights abuses and helping Semple make contacts for his work on political reconciliation in Afghanistan. Amanullah had originally fled to Pakistan, according to Semple, because one of the warlords who rose to power after 2001, Qazi Kabeer, had been a personal rival. But Semple said Amanullah had suffered such “an horrendous stint in ISI [Inter-Services Intelligence] detention” in Pakistan that he returned to Afghanistan to live openly in Kabul with his wife and children, and was in contact with pro-regime political figures. He had been so “psychologically bruised” by the ISI experience that he was incapable of working covertly for the Taliban, according to Semple.

But former BBC correspondent Kate Clark, who also knew Amanullah personally, has documented in detail how US intelligence analysts had convinced themselves on the basis of their analysis of cell phone traffic that Amanullah and the actual shadow Taliban Gov. of Takhar, Muhammad Amin, were the one and the same man. They had concluded that the Taliban shadow province chief for Takhar, Muhammad Amin, must have taken the name “Zabet Amanullah” as his alias. The analysts had not even done the most elementary checking to see if there was an actual Zabet Amanullah.

What sealed Amanullah's fate was that he was personally acquainted with Amin and had occasional phone contact with him. Semple, one of world's most knowledgeable specialists on the Taliban movement, located Muhammad Amin in Pakistan and interviewed him six months after the Americans had supposedly killed him. Amin convinced Semple that he was indeed the former shadow governor of Takhar and even showed him his identification card. Amin confirmed to Simple that he had spoken with Amanullah roughly once a month by phone.

Amanullah had gone to Takhar to help his nephew's parliamentary campaign. Nine other Afghans – all campaign workers for his nephew – were killed along with Amanullah in an SOF helicopter attack on two cars. He had cleared his visit to Takhar with the Shah Jehan, the police chief in Takhar, who was a personal acquaintance, according to Semple. A simple phone call to Jehan would have easily disabused the SOF of its assumption about Zabet Amanullah as the alias of the Taliban province chief.

The idea that the Taliban governor of Takhar would be riding along with the campaign staff of a candidate for the Afghan Parliament in broad daylight is so obviously implausible that it should have been a flashing red light for the operation. But Petraeus's International Security Assistance Force (ISAF) was immune to such logic. It issued a statement more than a week after the attack that suggested those riding along with the man who was killed must also have been insurgents. “The question remains,” the statement said, “why an election official or candidate was travelling with a known terrorist.”

Even after Afghan government officials had unanimously condemned the killing of Amanullah as a horrible mistake, Petraeus told PBS “Frontline” at the end of January that the killing was “a very precisely targeted operation” and that “there is no question about who this individual was.” Petraeus was so confident that they had gotten the right man that he agreed to Clark's request to meet with officers of the SOF unit involved in the murder of Amanullah.

It was apparently the first time that SOF personnel had ever talked about a specific raid with anyone outside the chain of command. The officers who met with Clark over dinner did not admit to having made a mistake in the killing, despite her presentation of all the evidence to the contrary. But when she pressed them about the evidence of the life and death of the actual Zabet Amanullah, they admitted something far more damning: that they had not been tracking Amanullah by name, but had only a cell phone number.

Guilt by Association

As for the nine campaign workers killed, one of the officers told Clark, “If someone is a targeted individual or someone with that person, they are unlawful combatants.” And a second officer said, “If we decide he's a bad person, the people with him are also bad.” Such statements are clearly at odds with the criteria used in humanitarian law to distinguish between combatants and civilians. And they shed light on the criteria used by the SOF commanders in tallying the number of “insurgents” they claim to have killed or captured.

In December 2010, ISAF gave pro-war blogger Bill Roggio impressive figures suggesting that the Taliban had lost 4,100 through capture and 2,000 were killed in the previous six months. Similar figures were released to selected journalists every three months after that. But an investigation of those figures reveals that all but a very small proportion of the total of insurgents detained in targeted raids were actually innocent civilians.

The publicly announced US detention policy as of 2010 was that all Afghans picked up in the field by US troops must either be sent to the Parwan detention facility or be released within two weeks. An unclassified February 5, 2011, internal document of the Combined Joint Inter-Agency Task Force responsible for detention policy in Afghanistan, which this writer obtained last spring, showed that only 690 Afghans were admitted to the US detention facility at Parwan during that six-month period – just 17 percent of the 4,100 captured insurgents claimed. The remaining 83 percent of those said to have been captured “Taliban” had actually been released within a few days because there was no evidence that they were indeed insurgents.

The proportion of civilians in the 4,100 captured “Taliban” was actually even higher than that. The same inter-agency detainee task force document shows that 20 percent of all those detained in Parwan during the six-month period were released upon further review of their files. So, the maximum number of detainees for whom there was any real evidence of active involvement with the insurgents was actually 552, or 14 percent of the total 4,100 captured “insurgents” claimed.

Many of the thousands detained in JSOC targeted raids were picked up because they happened to be present in a house that had been targeted. But many others, like Amanullah, were targeted because their mobile phone number had shown up too many times on a “map” of an insurgent network. Innocent civilians are much less likely to resist violently when their houses are raided by SOF, so the proportion of those killed in night raids who are civilians is bound to be much lower than the proportion of civilians among those who are detained, at least temporarily.

Hiding Civilian Deaths in Night Raids

But that does not mean that civilians represent a small proportion of the deaths in night raids. SOF units have routinely hidden such civilian deaths by reporting them as insurgents – even when it was perfectly obvious that they could not have been combatants. A raid on a compound in Gardez on February 12, 2010, killed two men, one of whom was a local government prosecutor and the other a senior intelligence official in the Afghan national police, and three women, two of whom were pregnant. But the SOF unit reported to the headquarters of the US-NATO command in Kabul that the two men were insurgents and claimed that the women had been found tied up and gagged. McChrystal defended the unit against the charges by eyewitnesses that its members had tried to cover up the killings of the three women even after the head of the Afghan Interior Ministry investigation of the incident publicly called the eyewitness testimony credible. Meanwhile, the target of the raid, who turned himself in for questioning a few days after the raid, was released without charge, according to the brother of the two men killed in the raid.

The United Nations mission in Afghanistan issued a report last March offering the figure of 80 civilians killed in what it called “search and seizure operations” in 2010, but that was a figure that it knew was highly misleading. Nader Nadery, a commissioner of the Afghanistan Independent Human Rights Commission, told my Afghan colleague Noori Shah Noori that that figure represented only the number of civilian deaths from13 incidents that had been fully investigated. It excluded the deaths from 60 other incidents in which complaints had been received, but had not yet been thoroughly investigated.

Nadery has since estimated that the total civilian deaths for all 73 night raids about which it had complaints was 420. But the Afghan Independent Human Rights Commission (AIHRC) admits that it does not have access to most of the districts dominated by the Taliban and that people in those districts are not aware of the possibility of complaining to the Commission about night raids. So, neither the AIHRC nor the United Nations ever learn about a significant proportion – and very likely the majority – of night raids that end in civilian deaths. The implication of that fact is that the majority of the more than 2,000 Afghans said to have been killed by SOF raids in 2010 may well have been innocent civilians rather than insurgents.

Two senior US commanders freely admitted to The Washington Post's Dana Priest and William Arkin that they had not targeted the right home or individual on more than “about 50 percent” of the raids. Given the tendency of commanders to overrate the success of their operations, that admission further underlines the vast human cost of the uncontrolled violence carried out by SOF units in Afghanistan.

Inviting Revenge for the Killing Machine

Afghanistan is the last place on earth the US military should be allowed to practice such indiscriminate killing. Afghans of every political stripe and at every level of society, from the Taliban fighters to President Hamid Karzai himself, have been warning the United States that the killing of innocent civilians in night raids provokes nearly universal hatred of Americans. And Pashtun culture is based on the necessity to take revenge against those who have harmed one's friends or family – even if takes the rest of one's lifetime.

Yet, the Obama administration, in seeming indifference to this well-established reality, is set to sign an agreement with the Karzai administration that will keep thousands of SOF in Afghanistan at least until 2014. Even worse, the program of targeted raids practiced in Afghanistan is now being touted as the military tactic of choice for future US wars. It is now clear that this “industrial strength killing machine” needs be brought under control and held accountable for the grave damage it has done to Afghans and to the interests of the United States.

We have hours left to raise $12,000 — we’re counting on your support!

For those who care about justice, liberation and even the very survival of our species, we must remember our power to take action.

We won’t pretend it’s the only thing you can or should do, but one small step is to pitch in to support Truthout — as one of the last remaining truly independent, nonprofit, reader-funded news platforms, your gift will help keep the facts flowing freely.