Testimony to the Tom Lantos Human Rights Commission, September 2026

Chairman McGovern, Chairman Smith, and Members of the Committee, thank you for inviting me to testify in today’s hearing. I have worked on matters related to democracy, human rights, and national security for over two decades with a more recent research focus on the impact of AI technologies on political repression, armed conflict, and governance.

I believe we are at a critical inflection point, where a combination of ever more powerful AI systems, accelerating proliferation, and a corresponding norms vacuum is setting up a dangerous set of conditions.

Military AI systems are no longer theoretical. Driven by developments on the battlefield in Ukraine, Gaza, and Iran, major militaries are rapidly integrating AI targeting systems at scale and experimenting with autonomous weapons. In the Iran war, during the first week of hostilities alone, the Pentagon reported it had struck more than 3,000 targets. By the time a ceasefire was announced 38 days later, the number of attacked targets had risen to 13,000, largely due to Palantir’s Maven targeting system. In Israel, AI targeting systems have played a crucial role in supporting its military campaign against Gaza. One of the deployed systems, known as Lavender, reportedly produced 37,000 target recommendations in the opening weeks of the war, leading to 15,000 strikes. In the Russia-Ukraine war, both sides have widely integrated AI into their targeting efforts. In late August, a small Russian drone swooped towards a Ukrainian gas station, exploded into a wall, and killed three civilians. Powered by a commercially available Nvidia minicomputer, and flying without a radio link, it reportedly used onboard AI to independently navigate and select its targets, making this one of the first documented cases of a fully autonomous, self-targeting AI system killing civilians.

As military applications of AI grow, their proliferation is increasing. Only a small number of countries currently possess high-end AI targeting systems equivalent to Maven: China, Israel, North Korea, Russia, South Korea, Ukraine, the United Kingdom, and the United States. These systems are complex and require extensive capital investments and infrastructural upgrades, limiting how quickly other nations can acquire them. Still, there is a significant pool of countries that possess the capacity and incentives to acquire this technology soon, such as India, Iran, Pakistan, Saudi Arabia, Türkiye, and the UAE—some of which have documented records of human rights violations and war crimes.

Other AI capabilities can also lead to wartime harm. For example, Anthropic’s September threat intelligence report revealed that a slew of actors used Claude Code to enhance their weapons development, including non-state actors like the Houthi militia, which deployed Claude to develop multi-stage ballistic missiles and guided rockets. And it wasn’t just the Houthis, reportedly an Iranian-linked group deployed Claude to target US Navy ships in the Middle East, pulling together transponder data, photographs and satellite imagery to track naval vessels. These developments are accelerating the diffusion of military knowledge and expertise, allowing hostile actors to achieve breakthroughs, such as developing autonomous drones, while facilitating the wide dissemination of lethal capabilities.

These developments put pressure on an already weakening global commitment to international humanitarian law (IHL) and international human rights. Not only is the world experiencing a surge in armed conflict—last year more than 255,000 people were killed in battle across 65 conflicts in 35 countries, the highest number of conflict events since 1946—but the scale of atrocities committed by warring parties is rising correspondingly. As the brutality of contemporary wars come into focus, there is scant political will to hold perpetrators to account. The United States has traditionally been a leading upholder of these principles, but under the current administration, America has executed a stunning reversal. In the Pentagon, Secretary Pete Hegseth fired top lawyers for the military services, known as judge advocates general (JAGs), who oversee military conduct and make sure that military personnel follow the laws of armed conflict. He eliminated offices entrusted with preventing and responding to civilian harm during combat operations, firing 90 percent of the employees staffing the Civilian Harm Mitigation and Response office and the Civilian Protection Center of Excellence. That decision has led to major deficiencies when it comes to evaluating civilian risks from drone and missile strikes and related missions. Reporting indicates, for example, that no member of either center reviewed the Minab site in Iran prior to US missile strikes that killed at least 178 civilians. Secretary Hegseth also fired the interim director and slashed personnel in the office of operational test and evaluation, which Congress established in 1983 to respond to concerns that the Pentagon was deploying weapons systems that failed to operate safely or effectively—a particularly troubling development in the age of AI warfare. Meanwhile, the Pentagon is pushing to incorporate AI technologies, from autonomous drones to AI targeting systems, into its arsenal, signaling its intent, in the words of Secretary Pete Hegseth, to become an “‘AI-first’ warfighting force across all components, from front to back.” 

Responding to military AI risks

The challenges presented by military AI are acute, but I believe that progress is possible and would offer four ideas for consideration:

• Advance legislation to codify requirements in the Pentagon for meaningful human judgment and accountability. Military personnel need to better understand how these systems work, and they should exercise appropriate oversight and judgment consistent with US law, the laws of armed conflict, and the military’s rules of engagement. When the Pentagon deploys an AI targeting system like Maven in war, it should collect data about how that system has performed—where did it succeed and in which instances did it commit errors and why. Without this information, the military is unable to accurately assess whether its strikes are hitting their intended targets, undermining military leaders’ ability to learn from mistakes and improve future operations. Congress can help address these shortcomings by enacting legislation.
• Support international negotiations on military AI. There is growing momentum to rein in these systems; on September 5, 128 states parties to the Convention on Certain Conventional Weapons (CCW) agreed to address lethal autonomous weapons and develop rules for meaningful human control over these systems. The next critical juncture comes at the November CCW Review Conference, where states will decide whether to negotiate binding rules. Congress should urge the US government to support, not obstruct that effort.
• Hold private tech companies accountable. Firms like Palantir, Anthropic, AWS, Google, and OpenAI are responsible for designing, developing, and operating complex military AI systems. As such, Congress should subject them to public scrutiny, including hearings and member briefings, to determine how these systems function and whether they are eroding human oversight in risky and unaccountable ways.
• Address the proliferation risks of military AI. A top priority for policymakers should be limiting the diffusion of military AI technology. For countries with a record of human rights violations, Congress should press the US administration to prevent these governments from acquiring AI targeting systems and its components, including surveillance technologies, compute infrastructure, and targeting software. Further, Congress should push the administration to develop rules preventing adversarial states and militant groups from accessing frontier AI models like Claude Code for weapons development.

To conclude, unconstrained innovation does not lead to better or more efficient warfighting. Rather, it produces unreliable weapons that miss their targets, violate human rights, and ultimately move the United States further from victory. The proliferation risks associated with military AI open the door to mounting civilian harm and threaten America’s national security. We have a narrow window of opportunity to shape this technology. We should seize this moment while we can.

Thank you. I look forward to your questions.