Pentagon sees broader role for AI in setting military targets

NEW YORK – The Pentagon has quietly revised its doctrine on how the US military picks its targets in battle, opening the way for artificial intelligence (AI) to make critical wartime decisions in the future.

The revised targeting principles, approved without public disclosure in April, envision “systems where AI initiates actions with human monitoring”, evolving from what it describes as the current practice of having “‘human in the loop’ systems, in which a human initiates actions”.

Bloomberg News reviewed the latest changes, which have not been previously reported. While not classified, the document is not released publicly.

The revised targeting guidance is the latest example of the Pentagon’s rush to accelerate adoption of AI, a technology that is already transforming battlefields in conflicts around the world. AI’s rapidly expanding capabilities are also raising new threats, as well as ethical dilemmas, challenging the military’s ability to keep pace and adhere to the law of armed conflict.

“The speed of future warfare, along with our adversaries’ own advances in AI, may require the joint force to adopt completely autonomous systems,” the publication says.

The controversial new vision is included in a new chapter on the future of targeting. The publication as a whole lays out official procedures for so-called joint targeting – how the US armed forces determine what to fire at in combat.

It offers new content on “civilian harm mitigation” just weeks after the Pentagon announced an investigation into a strike on an elementary school beside an Islamic Revolutionary Guard Corps naval base in the Iranian city of Minab that killed an estimated 120 children.

It is not clear whether the document reflects lessons from the Iran war; changes to doctrine are typically lengthy undertakings. The previous update was made in September 2024, before US President Donald Trump returned to office.

The Pentagon did not respond to a request for comment.

Earlier on June 25, the Pentagon announced it would field advanced, AI-enabled tools to “compress the time it takes to turn intelligence into informed options for commanders around the world”, saying it would transform battle management, decision support, and targeting.

The system would not autonomously select or strike targets, but ensure commanders have access to options based on continuous scanning of defense intelligence and operational systems, the Pentagon said.

While previous editions of joint targeting doctrine have been published online, sometimes as a result of requests under the Freedom of Information Act, nothing beyond the outdated 2018 version is currently available to the public.

The newly revised document, the most explicit indication to date of the Pentagon’s focus on removing humans from life-and-death decisions, makes clear that AI will reduce the “sensor to shooter cycle” and “increase the tempo of operations”.

At the same time, the new doctrine recognises the risks of relying on AI exclusively, saying that it “poses serious moral and legal dilemmas and requires the establishment of unambiguous ethical guidelines to mitigate concerns with AI enhanced decisions”.

The updated doctrine does not introduce that guidance, but a new national security presidential memorandum released earlier in June gave the Pentagon 90 days to update its policy on autonomy in weapons systems.

At the moment, the policy requires autonomous and semi-autonomous weapons systems to undergo reviews to ensure they are designed to allow commanders and operators to exercise “appropriate levels of human judgment over the use of force”. That stops far short of aims from campaigners who seek a stronger safeguard of meaningful human control.

UN Secretary-General Antonio Guterres is among those who have called for a ban on lethal autonomous weapons systems, calling them “politically unacceptable” and “morally repugnant”.

Anthropic, which provides its AI tools to the Pentagon’s classified networks, including for use in Iran operations, fell out with the agency over the company’s insistence that the technology is not yet reliable enough for what chief executive Dario Amodei described as “fully autonomous weapons” and that humans should make the final decision to use force.

The new targeting doctrine for its part says the military “must exercise caution to avoid over-reliance on AI output without human oversight”, noting that AI algorithms can ingest and analyse intelligence products “at a scale and speed far exceeding human capacity”.

It also outlines a larger role for AI in improving the quality of targeting data, helping to close gaps in intelligence and automate data flows.

Several targeting experts have expressed hope that connecting up digital systems and introducing more AI – not less – could nevertheless help address and fix targeting problems in future. That might include, for example, making an automated check against public sites, such as Google Maps, to help flag any potential anomalies for human review.

“Advancements in AI will improve the targeting process and enable long range precision engagements,” the Pentagon document reads, saying commanders will “need to harness the power of AI” to augment human analysis, decision making and risk management, in order to maintain the joint force’s advantage.

“AI may quickly compare collected target data from multiple platforms to corroborate veracity of collected information,” the document continues.

But the latest joint targeting document also adds caution in an appendix devoted to the historical “challenge” of integrating automation into targeting, saying that “automation is not a replacement for human thinking or proactive communications”.

Regardless of any automated tool used, it says “commanders are responsible for the priorities, effects, and timing” within their operational area and for the adherence to the law of war and relevant US military rules of engagement. BLOOMBERG