Pentagon Pushes Battlefield AI as Leaders Urge Caution

The Trump administration is accelerating the integration of artificial intelligence into U.S. military operations, sparking a multi-front debate that pits Defense Secretary Pete Hegseth’s push for unrestricted AI use against senior military leaders who insist humans must remain in control of lethal decisions. The tensions have erupted into public view through a bitter legal battle between the Pentagon and AI company Anthropic, which refused to allow its technology to be used for autonomous weapons or mass domestic surveillance.

The Central Conflict: Speed vs. Oversight

According to AP News, Adm. Frank Bradley, head of U.S. Special Operations Command, told attendees of the SOF Week conference in Tampa, Florida, that troops “have to be very careful about how we come to (AI’s) employment and its inspiration into the delivery of lethality.” Bradley said he can envision a future where AI helps determine targets but insisted that “we, as humans, have to have the confidence that … it’s going to deliver violence only where we intend it to be delivered.”

Bradley’s caution stands in direct contrast to his boss, Defense Secretary Pete Hegseth, who has insisted the Pentagon be allowed to use AI technology “any legal way it sees fit.” Hegseth told an audience of SpaceX employees in January that he would reject any AI models “that won’t allow you to fight wars” and that his vision was systems operating “without ideological constraints that limit lawful military applications.”

The Anthropic-Pentagon Showdown

The debate has crystallized around a high-stakes confrontation between the Pentagon and Anthropic, the San Francisco-based AI company behind the Claude chatbot. Anthropic CEO Dario Amodei refused Pentagon demands to allow Claude to be used for fully autonomous lethal weapons and mass domestic surveillance without judicial authorization. In response, the Pentagon formally labeled Anthropic a “supply chain risk,” ending a $200 million defense contract and prohibiting other government contractors from working with the company.

Anthropic sued the Pentagon in March 2026, filing a 48-page lawsuit in the U.S. District Court for the Northern District of California. The company argued that “the Constitution does not allow the government to wield its enormous power to punish a company for its protected speech.” The case has drawn amicus briefs from Microsoft, retired military chiefs, and employees of Google and OpenAI.

White House spokeswoman Liz Huston defended the administration’s position, stating: “Under the Trump Administration, our military will obey the United States Constitution — not any woke AI company’s terms of service.”

AI Already on the Battlefield

While the policy debate continues, AI is already being deployed in active combat operations. Lt. Gen. Michael Conley, head of Air Force Special Operations Command, told a congressional committee in May 2026 that his troops used AI “bots” to convert top secret intelligence to a secret classification within seconds during the ongoing U.S.-Iran war, dramatically speeding up information sharing with drone operators on the ground.

A Georgetown CSET case study documented how the Army’s 18th Airborne Corps used AI to target artillery strikes “just as efficiently as the best unit in recent American history” while operating with 2,000 fewer service members. Helen Toner, interim executive director of Georgetown CSET, noted that “human operators are still the ones making crucial decisions, but AI … is making it possible to operate with a new level of speed and scale.”

Diverging Visions Within the Military

According to Orbital Today, Bradley’s remarks highlight an ongoing debate within the defense establishment over the appropriate boundaries for machine decision-making in combat. While the Pentagon has invested heavily in AI for intelligence, surveillance, and reconnaissance, senior special operations leaders see AI differently than the civilian leadership.

Sgt. Maj. Andrew Krogman, the top enlisted official for U.S. Special Operations Command, said at the SOF Week conference that he sees AI handling administrative tasks to free up operators. Melissa Johnson, the top acquisition official for the command, said AI should focus on “reducing the cognitive workload on mundane tasks” rather than replacing operator judgment.

Toner, a former OpenAI board member, said “the general public often seems to underestimate the caution with which the U.S. military approaches new technologies,” adding that “commanders want their missions to succeed, which means both being able to create lethal effects at scale, and avoiding unintended effects like friendly fire, civilian casualties, or simply identifying targets incorrectly.”

Tech Worker Activism Resurfaces

The dispute has also reignited tech worker activism not seen since Google employees protested Project Maven in 2018. Over 200 employees at Google and OpenAI signed an open letter demanding their companies adopt the same red lines as Anthropic — no autonomous lethal weapons and no mass domestic surveillance. However, the political environment has shifted significantly: in 2026, the White House is actively pressuring tech companies to cooperate with military AI development, making internal dissent more costly.

The Canceled AI Executive Order

President Trump abruptly called off plans to sign a new AI executive order on May 21, 2026, hours before an expected White House ceremony. The draft order would have created a voluntary 90-day federal review process for frontier AI models, prompted partly by concerns over Anthropic’s advanced capabilities. However, after pushback from AI adviser David Sacks and other Silicon Valley allies who warned it could slow U.S. competitiveness with China, Trump pulled the order.

“We’re leading China, we’re leading everybody, and I don’t want to do anything that’s going to get in the way of that lead,” Trump told reporters.

What to Watch For

The outcome of Anthropic’s lawsuit against the Pentagon could set a precedent for how the government contracts with AI companies. The case has produced conflicting rulings at different court levels — a federal judge blocked the Pentagon’s supply chain risk designation in March, but a Federal Court of Appeals denied Anthropic’s motion to lift it in April. Meanwhile, the U.S. lacks a unified framework for frontier AI governance, while the EU has binding rules under the AI Act, leaving a policy vacuum as competition with China accelerates the race for military AI dominance.