On the first day of a military campaign against Iran, a strike by United States forces resulted in the deaths of 123 children at an elementary school. The strike was facilitated by the Maven Smart System, an AI-enabled software platform developed by Palantir Technologies, which aggregates and processes battlefield data to identify targets. Though the elementary school had been converted from a naval base in 2017, the intelligence fed into the Maven system was outdated, leading the AI to mark the school as a priority military target.
The incident, described by some as one of the most significant massacres involving U.S. forces in decades, has exposed a critical legal vacuum in international law: the “diffusion of responsibility” when humans and machines collaborate on lethal decisions. As autonomous and AI-assisted weapons move from theoretical concepts to active battlefield tools, the global community is struggling to determine who should face prosecution when these systems violate the laws of armed conflict.
In the wake of the strike, a clear divide has emerged between the military and the private sector. Officials within the U.S. Central Command have expressed that personnel expected the Maven system to flag inconsistencies or stale data within the underlying intelligence. In contrast, Palantir Technologies has maintained that it is not responsible for the underlying data provided to the system, nor is it responsible for identifying deficiencies in intelligence. The company has denied that its software was at fault, pointing instead to the human-provided data sets it was tasked with processing.
International law expert Allen Weiner, a senior lecturer at Stanford Law School and former State Department lawyer, suggests that the current legal framework is ill-equipped for this new reality. Traditional understandings of responsibility usually follow two paths. In one theory, if an illegal attack occurs due to faulty code or programming, the manufacturer could be held liable. In another, responsibility lies with the commander who decides to deploy a tool capable of making autonomous judgments, even if they did not know the specific outcome of the deployment in advance.
However, the requirement for a “culpable mental state”—the intent or knowledge necessary to convict someone of a war crime—presents a high bar. If a commander gives an order to “hit the enemy hard” and the AI interprets this as a mandate to attack civilian targets like schools to maximize impact, the commander may argue they lacked the specific intent to target children. If the developers did not intend for the machine to make such a choice, the blame is diffused across a web of programmers, intelligence analysts, and officers, making it nearly impossible to identify a single individual with the required criminal mindset.
While identifying a “war criminal” may be difficult, Weiner notes that state responsibility remains a well-established principle. Under international humanitarian law, the country that carries out an attack against civilians is responsible for the violation, regardless of whether a machine or a human made the final call. This state-to-state liability often manifests as financial reparations or claims brought before international tribunals, such as the UN Compensation Commission used after the first Gulf War.
Beyond the military chain of command, the role of private tech firms is also coming under scrutiny. There is an emerging debate over whether companies like Palantir or Anthropic could eventually be classified as combatants under international law. While a factory producing airplanes has long been considered a legitimate military target, the status of a company providing the digital brains for autonomous warfare is less clear. If private software is directly participating in hostilities by making targeting decisions, the personnel of these companies may lose their civilian immunity.
As the line between human and machine decision-making continues to blur, the prospect of “war torts” has been proposed as a middle ground. This would allow victims to sue for negligence or reckless harm in civil courts, bypassing the high evidentiary standards of criminal war crimes trials. However, such a shift would require significant legislative action to overcome hurdles like federal sovereign immunity, which currently protects government agencies and many of their contractors from private litigation.
