A recent news story by Glenn Greenwald and Jeremy Scahill details the use of NSA signals intelligence (SIGINT) – including cellphone and SIM card data – to locate and kill suspected militants in Afghanistan, Iraq, Pakistan, Somalia, and Yemen. It has long been public knowledge that US operations use mobile phone SIGINT in this way to carry out military strikes (since at least 2004)—and that so do our allies such as Canada (since at least 2004) and Israel (since at least 2003).  So what’s new here? The key revelation in the Greenwald and Scahill story is that the United States may be increasingly dependent on such SIGINT and that this form of intelligence can have serious reliability problems in some situations–with the result that the wrong people may be killed.

We have our own independent criticisms of the Greenwald and Scahill report (see section II below). However, the response by two highly esteemed former US military lawyers – Geoffrey Corn and Chris Jenks — is overly dismissive. And, their commentary may create the false impression that the law of armed conflict (LOAC) is indifferent toward the moral concerns raised by such knowing use of faulty intelligence in lethal targeting.

On the contrary, LOAC has a lot to say about the use of unreliable intelligence in military strikes. And, if Greenwald and Scahill’s reporting is accurate, there is a risk that some US combat operations may be in violation of LOAC and US policy directives to avoid civilian casualties.

[As an aside: the reported US actions would, a fortiori, violate international human rights law if these operations occur outside an armed conflict – a point with which we believe Corn and Jenks would agree.]

Importantly, regardless of what the law requires, the President has called for minimization of civilian casualties and targeting only specific individuals—for moral and strategic reasons. Military operations that persist despite preventable and correctible errors would presumably also run counter to such policy dictates.

I. Precautions in attack

1. What legal rules apply?

LOAC requires that in the conduct of military operations “constant care shall be taken to spare the civilian population” and that “those who plan or decide upon an attack shall…do everything feasible to verify that the objectives to be attacked are neither civilians nor civilian objects” (our emphasis added). As the US military’s Law of Armed Conflict Deskbook explains, this rule means that commanders must take “all practicable precautions” to avoid loss of civilian life.

We thus agree with Corn and Jenks’s statement that there would be no culpability on the part of US commanders under LOAC “if the U.S. has undertaken reasonable precautions leading up to the strike” (emphasis by Corn and Jenks). The important legal question then is whether the US is abiding by all practicable precautions in attack, taking into account the relevant circumstances and the information available at the time. And this is, in large part, the question to which Greenwald and Scahill’s piece is directed. Their reporting suggests that the US too readily equates the location of certain cell phones or SIM cards with the location of a suspected terrorist.  If the US has information – as one of the drone operators in the piece suggests – that terrorists are known to give their cell phones to civilians as an evasive tactic, the US has a legal obligation to take that information into account during attack, and to take reasonable precautions in light of it.

What about the responsibility of the militants who are purposefully risking the lives of innocent civilians? Corn and Jenks overstep when they later write that taking into account the responsibility of the terrorists “supports a fundamentally different conclusion”:

“it is the terrorist operative who uses and then gives his cell phone to a civilian in an effort to avoid being the object of US attack, who is both legally and morally culpable for the death of that civilian, not the US commander” (emphasis added)

Corn and Jenks are wrong on the law if they mean to suggest that the actions of the terrorist operative absolves the US of its obligation to show “constant care” for civilian life and exhaust practicable measures to avoid killing civilians.

We agree with Corn and Jenks that an individual taking part in hostilities who intentionally co-locates civilians with a military target is violating basic morality and binding legal rules. Indeed, we would take it one step further than Corn and Jenks do in their post: on our view, such actions by militants could constitute a war crime in non-international armed conflict. Depending on the circumstances, intentionally distributing cells phones to civilians for the purposes of evading attack could breach the prohibition against the use of human shields, and/or the principle of distinction and the obligation to take feasible actions to protect civilians against the effects of attacks.

That said, LOAC could not be clearer on the continued responsibilities of the attacking party: “Any violation of these provisions [including the use of civilians to shield combatants or impede an attacker’s military operations] shall not release the [attacking party] from their legal obligations with respect to the civilian population and civilians, including the obligation to take the precautionary measures”  (Article 51(8) of Geneva Protocol I).

In short, Corn and Jenks take Greenwald and Scahill to task for “ignor[ing] the legal and moral responsibility the terrorists bear for civilian deaths from SIGINT based drone strikes.”  But in so doing, Corn and Jenks fail to adequately lay out the US government’s own legal obligations when it launches attacks. It is well-settled law that both parties have legal obligations during attacks. The use of human shields – as unlawful and atrocious as it is – does not absolve the other (attacking) party of its own legal obligations.

2. How should we evaluate the reliability of cellphone location-based targeting?

Corn and Jenks sensibly point out that targeting has often relied on SIGINT in a variety of contexts. They also make a good case that SIGINT is often the best information available in the targeting decision-making process and that NSA SIGINT often provides accurate information on the identity, associations, and activities of specific cellphone users.

All that said, though, the NSA SIGINT is potentially unreliable for targeting decisions when used in the specific manner described in Greenwald and Scahill’s report. They describe the use of this SIGINT in two ways: (1) it is used to link a suspected enemy to a particular cellphone (or SIM card); and (2) it is used to track the location of that cellphone (or SIM card). As reported by Greenwald and Scahill , the U.S targets the location of the cellphone (or SIM card) for a lethal strike. The question, then, is whether the location of the cellphone (or SIM card) is a sufficiently reliable proxy for the location of the suspected “enemy combatant.” Proxies of a similar sort are often somewhat reliable—and, at least, somewhat useful in targeting decisions. Individual terrorists might be linked to a particular computer, automobile, house, or local establishment. These proxies, like the cellphone, share certain characteristics that undermine their reliability. The problem, of course, is that they can be temporarily or permanently passed to civilians or abandoned altogether. Given the well-known practice of terrorist groups of sharing, re-assigning, or abandoning cellphones (or SIM cards), this source of unreliability is only accentuated. At bottom, these proxies may be reliable at a certain point in time when the intelligence is gathered, but their value degrades over time—and this might happen very quickly if the intended target is aware of the use of the proxy. This is not to say that this information is useless—not at all. The point is instead that this information typically would need to be coupled with other intelligence indicating the location of the target and the circumstances surrounding the target at the time of the strike/no-strike decision. Consider the following account of Israel’s practices which include but do not rely on cellphone SIGINT:

“Israel has used smart rockets guided toward the electromagnetic signals emanating from mobile phones …

Military analyst Naji al-Batta told Al-Monitor that Israeli intelligence relies on both technology and human resources to obtain its information. ‘Israel cannot hit the resistance unless it is sure of the target. The Israeli units recruit human resources to provide the intelligence….’”

3. How should we evaluate the error cost if cellphone-based targeting results in the unintended killing of other militants?

That question may, at first blush, seem like it has an obvious answer. Indeed, Corn and Jenks assert with confidence: “that an unintended Taliban leader was killed is squarely in the ‘not a problem’ category.” Their construction of a scenario in which a different Taliban leader is killed is revealing. The Greenwald and Scahill report is not limited to such unintentional targets. Instead the alleged reliability problems of cellphone and SIM card location technology may result in other members of the armed group being killed. But if they are combatants, who cares?

The President does. The Obama administration for sound legal and strategic reasons targets only senior members of Al Qaeda, the Taliban, and associated groups in many if not most situations. Knowingly conducting targeting operations that have a significant error rate in killing other members of the group may violate the purpose and substance of those directives.

Moreover, if the target ends up being a low-level member of the Taliban, it could fundamentally affect the proportionality analysis under LOAC. For example, it may be lawful to kill a Taliban leader even though several civilians are incidentally injured in the operation—but not if it’s a low-level Taliban member.

Finally, why limit this question to the Taliban? In the context of targeting members of al-Shabaab, for example, the US has been careful to distinguish between leaders who support and those who oppose striking the West. Indiscriminate targeting operations would fly in the face of the wisdom of such restraints.

II. Our own concerns about the Greenwald and Scahill report

As we mentioned at the outset, we naturally have our own concerns with the Greenwald and Scahill report.

First, the story implies that cellphone-based targeting and overreliance on SIGINT are the problem—solved by more HUMINT. But a solution to bad SIGINT may be more SIGINT in some circumstances. At least that option should not be excluded. What might the “enhanced SIGINT” option include? One example: Greenwald and Scahill report that the NSA often locates targets “by analyzing the activity of a SIM card, rather than the actual content of the call” (emphasis added). In that event, one solution worth exploring in warfighting is collection and analysis of the content of the call before ordering a strike—to help make sure the correct person is on the other end of the line.

Another example of “enhanced SIGINT” options is in the more distant future.  As more miniaturized UAVs – e.g., the size of a hummingbird or an insect — are developed for surveillance purposes, SIGINT may overcome cellphone targeting problems. There may be many other sound reasons to reject the use of such technology (especially outside of warzones), but their increased accuracy should be considered as part of the policy calculation. For example, a recent academic paper, which is highly critical of the overreliance on SIGINT for UAV and signature strikes, concludes with the following acknowledgement:

“[M]icro UAVs may be able to eliminate or reduce the need for traditional human espionage. The UAVs will soon have the potential to float in place by a window while recording a conversation. They could also nonchalantly observe the layout of a small room under the disguise of a fly or cockroach. These usages are all in line with the original intention of UAVs, which was to avert the loss of human life when at all possible.”

Second, the Greenwald and Scahill story does not sufficiently consider contradictory evidence that undermines their claim that reliance on cellphone SIGINT is causing more civilian deaths. The authors make three claims in an effort to connect the dots:

  1. Reliance on cellphone SIGINT for drone strikes has been increasing.
  2. It is “especially” “more likely that mistakes are made” using this approach in Pakistan.
  3. “The Bureau of Investigative Journalism, which uses a conservative methodology to track drone strikes, estimates that at least 273 civilians in Pakistan, Yemen and Somalia have been killed by unmanned aerial assaults under the Obama administration.”

Here’s a problem with the above: At the same time that cellphone tracking for drone strikes has been increasing, the civilian casualty rate in Pakistan drones strikes has been reducing — according to the very same Bureau of Investigative Journalism. Indeed, according to their annual report, there may have been zero civilian deaths in Pakistan in the past twelve months. Greenwald and Scahill’s aggregate statistic – of 273 civilian deaths in Pakistan, Yemen, and Somalia combined over the past five years – masks these recent trends, which are inconsistent with their account and, on the contrary, more consistent with the idea that cellphone SIGINT might improve targeting accuracy.

* * *

In sum, Greenwald and Scahill raise important questions that may require a fundamental rethinking of existing efforts to integrate specific elements of NSA SIGINT into US kill or capture operations. The evidence they provide, admittedly of a preliminary character, suggests concrete legal and policy concerns that demand sober reflection.