Trump Bans Anthropic as OpenAI Wins Pentagon Deal

Anthropic, currently valued at $380 billion, has held a $200 million Pentagon contract since July 2025 and is the only AI company with models deployed on classified military networks, operating through a partnership with defense technology firm Palantir. By all accounts, the working relationship had been productive — until the Defense Department demanded that Anthropic agree to "any lawful use" of its AI tools without restriction.

Anthropic pushed back, requesting contractual language enshrining two specific commitments: that its technology would not be used for mass surveillance of Americans, and that fully autonomous weapons systems would retain human oversight. The company maintained that these were not radical demands but rather codifications of principles already embedded in U.S. law and longstanding Defense Department directives. According to [reporting by the BBC](https://www.bbc.com/news/articles/cn48jj3y8ezo), a former Department of Defense official said Anthropic held the "upper hand" in negotiations, noting the company had "great PR" and "simply do not need the money."

The Pentagon countered by offering what it described as a written acknowledgment of existing laws that already restrict surveillance and autonomous weapons deployment. Anthropic rejected this, saying the new language "made virtually no progress" and contained "legalese that would allow safeguards to be disregarded at will." The gap between the two sides was not merely semantic — it reflected a fundamental disagreement about whether contractual safeguards should be enforceable commitments or acknowledgments of existing legal frameworks that could be reinterpreted or waived. Defense Secretary Hegseth then escalated dramatically, threatening to invoke the Defense Production Act and the supply-chain risk designation if Anthropic did not comply.

Anthropic CEO Dario Amodei responded that he would rather stop working with the Pentagon entirely than acquiesce to what he called threats. [CBS News reported](https://www.cbsnews.com/news/trump-anthropic-ai-order-federal-agencies/) that President Trump subsequently directed "EVERY Federal Agency" to "IMMEDIATELY CEASE" all use of Anthropic's technology, with the president also threatening "civil and criminal consequences" if the company fails to cooperate during a six-month phaseout period. The speed of the escalation — from contract negotiation to a government-wide ban — stunned observers across the technology and defense sectors.

Anthropic has long positioned itself as the safety-focused leader in the AI industry, and CEO Dario Amodei framed the dispute as a test of whether AI companies can maintain ethical principles under government pressure. In public statements, Amodei called the administration's actions "retaliatory and punitive," arguing that the company was being singled out for exercising its right to set conditions on how its technology is used.

The company announced it will challenge the supply-chain risk designation in court, calling it an unprecedented and legally questionable application of a national security tool against a domestic company. The designation, which has historically been applied to entities linked to foreign adversaries such as Chinese telecommunications firms, carries significant consequences — it can restrict a company's ability to do business with the federal government and signal to private-sector partners that the firm poses security risks.

Anthropic's stance rests on a straightforward argument: the two safeguards it requested — no mass surveillance of Americans and human oversight of autonomous weapons — are consistent with existing U.S. law and longstanding Department of Defense policy. The company contends that enshrining these principles in contractual language should have been uncontroversial, and that the Pentagon's refusal to do so raises troubling questions about the government's intentions.

Critics within the defense establishment see it differently. Chief Pentagon technology officer Emil Michael called Amodei a "liar" with "a God-complex," [according to NBC News](https://www.nbcnews.com/tech/tech-news/trump-bans-anthropic-government-use-rcna261055), and accused the company of seeking to impose its own judgment over democratically accountable government officials on matters of national security.

Just hours after the Anthropic ban was announced, OpenAI CEO Sam Altman revealed that his company had reached an agreement with what he referred to as the "Department of War" to deploy its AI models on the Pentagon's classified network. The timing immediately drew scrutiny, with observers noting the apparent connection between Anthropic's ouster and OpenAI's entrance. The juxtaposition was striking: one AI company blacklisted for requesting safety guardrails, another welcomed aboard the same day.

What made Altman's announcement particularly notable was his claim that OpenAI's deal includes the same two restrictions Anthropic had sought: no mass surveillance of Americans and no fully autonomous weapons without human oversight. [CNBC reported](https://www.cnbc.com/2026/02/27/openai-strikes-deal-with-pentagon-hours-after-rival-anthropic-was-blacklisted-by-trump.html) that Altman had earlier told OpenAI employees that his company shares the same "red lines" as Anthropic on these issues. The internal communication suggested OpenAI's leadership viewed the safeguards as non-negotiable ethical commitments rather than bargaining positions.

Altman went further, publicly asking the Department of Defense to "offer these same terms to all AI companies" — a statement that appeared to validate Anthropic's position while simultaneously positioning OpenAI as both a willing defense partner and a company that shares its rival's safety commitments. The move was interpreted by some analysts as a deft piece of corporate positioning: OpenAI gains access to one of the most lucrative and strategically important AI contracts in the world while publicly aligning itself with the safety principles that got Anthropic banned. Others saw Altman's public call for universal terms as a genuine act of industry solidarity that could ultimately benefit Anthropic's legal challenge.

The Pentagon also mentioned that Grok, the AI system developed by Elon Musk's xAI, could potentially be used in classified settings — a detail that further fueled speculation about the political dimensions of the dispute, given Musk's close relationship with the Trump administration. The suggestion that multiple AI providers were being considered for classified deployment underscored the strategic importance of the market that Anthropic had previously dominated alone.

The Anthropic ban has rapidly become a partisan flashpoint. Senator Mark Warner, the Virginia Democrat and ranking member of the Senate Intelligence Committee, accused President Trump and Defense Secretary Hegseth of "bullying" Anthropic into deploying "AI-driven weapons without safeguards." [Fox News reported](https://www.foxnews.com/politics/dems-potential-2028-hopefuls-come-out-against-us-strikes-iran) that multiple Democratic lawmakers have raised concerns about the precedent being set.

Defenders of the administration's position argue that the federal government must have unrestricted access to the best available AI technology for national defense purposes, and that private companies should not be able to unilaterally dictate the terms under which the military operates. They point to the Pentagon's offer to acknowledge existing legal restrictions as a reasonable compromise that Anthropic rejected.

The former Department of Defense official who spoke to the BBC offered a more nuanced assessment, suggesting that the confrontation was as much about power dynamics as policy substance. By the official's account, Anthropic's strong financial position — with a $380 billion valuation and no pressing need for government revenue — gave it unusual leverage in negotiations with the Pentagon, and the administration's heavy-handed response may have been driven in part by frustration at dealing with a company that could afford to walk away.

Legal experts have also weighed in on the supply-chain risk designation, with several noting that applying it to a domestic American company represents uncharted legal territory. Anthropic's decision to challenge the designation in court could produce a landmark ruling on the scope of executive authority over the domestic technology sector, particularly in the rapidly evolving AI space.

The standoff between the Trump administration and Anthropic carries significant implications for the broader artificial intelligence industry. At its core, the dispute raises a question that every major AI company will eventually have to confront: what happens when government demands for unrestricted access to AI technology collide with a company's stated ethical commitments?

For AI startups and established players alike, the Anthropic ban sends a clear signal about the risks of pushing back against federal customers. The supply-chain risk designation, if it survives legal challenge, could become a powerful tool for compelling compliance from technology firms — a prospect that has alarmed civil liberties advocates and industry executives. The six-month phaseout period and the threat of civil and criminal consequences add further pressure.

At the same time, OpenAI's successful negotiation of terms that apparently include the same safeguards Anthropic sought suggests that the administration's objection may have been less about the substance of the restrictions than about which company was setting the terms. If OpenAI's deal genuinely includes prohibitions on mass surveillance and autonomous weapons, it raises the uncomfortable question of why Anthropic was punished for requesting the same protections.

The mention of xAI's Grok as a potential alternative for classified work introduces another dimension: the growing intersection of political relationships and AI procurement. As the federal government becomes the single largest customer for AI services, the risk of political considerations influencing which companies receive contracts — and which are shut out — has become a pressing concern for the industry.

For the defense establishment, the immediate practical question is straightforward: Anthropic's models are currently the only AI systems deployed on classified networks, and replacing that capability during a six-month phaseout will be a significant technical and operational challenge. The longer-term question is whether this episode will discourage the best AI talent and companies from working with the government at all — a concern raised by observers across the political spectrum.