- OpenAI signed a deal with the U.S. Department of Defense to allow its AI models to run inside classified military networks, stepping in after rival AI company [Anthropic refused similar terms over concerns about surveillance and autonomous weapons.](@story/anthropic)
- The U.S. government responded to Anthropic’s refusal by banning its tools from federal use and labeling the company a “supply-chain risk,” effectively pushing it out of certain defense partnerships.
- OpenAI said its agreement includes safeguards such as limits on domestic surveillance and requirements for human responsibility in the use of force, though critics argue these protections are not fully legally binding.
- The deal triggered major backlash from the public, tech workers and AI safety advocates, with millions of users pledging to boycott ChatGPT and critics accusing the company of compromising AI safety principles.
- The dispute has expanded into a broader debate about the role of private AI companies in warfare and national security, highlighting tensions between AI safety commitments and governments seeking advanced AI capabilities.
A US temporarily blocks the Pentagon’s “” designation against Anthropic, allowing the company to continue working with contractors. Judge Rita Lin says the move appears punitive and possibly retaliation for Anthropic’s stance against mass surveillance and autonomous weapons. The ruling pauses enforcement while the case continues, marking the first legal check on government action in this dispute. This shifts the conflict from policy and debate into direct judicial intervention over control of military AI partnerships.
During ongoing hearings, Judge Rita Lin said the key issue is not just government procurement but whether the punished Anthropic for its views. She questioned if the "" label was used as retaliation for refusing and use. This shifts the case toward a possible violation of "" and "" rights. The court indicated that if the government used powers to target a company’s stance, it could set limits on how far agencies like the "" can go when dealing with private AI firms.
A federal brought the Anthropic vs Pentagon dispute into the spotlight as Judge Rita Lin openly questioned the "" label imposed on Anthropic. She suggested the move may be punishment for the company’s stance on AI safety, especially its refusal to allow use of Claude in mass surveillance or autonomous weapons. The judge called the government’s actions "troubling" and raised concerns that the blacklist could be an attempt to cripple the company rather than protect national security. The Pentagon defended the decision as necessary for reliability and wartime readiness, while Anthropic argued it violates free speech and due process. This moment shifts the conflict from a policy fight to a legal battle that could redefine control over military AI use.
Retired judges and ethical groups backed Anthropic in court, arguing that use of in and rses serious concerns about . The support expands the conflict beyond a tech dispute into a broader societal debate about how should be used in warfare. The case is now becoming a test of whether limits on military use of will come from companies, governments, or public pressure.
A major was scheduled to examine whether the Pentagon’s against Anthropic is legally valid. The hearing will test if the government can block companies that restrict of , or if firms can enforce limits on how their models are used. This case could set a precedent for how much control AI companies retain once their systems enter .
Anthropic told a U.S. court that it cannot interfere with once deployed by the , rejecting claims of a hidden “” or sabotage ability. The company said there is no “” or backdoor access, meaning control shifts to the systems running it. This directly challenges the Pentagon’s argument that Anthropic is a because it could interfere during wartime. The case now hinges on whether modern AI systems can realistically be controlled after deployment or if that fear is misplaced.
Nearly 150 submitted supporting Anthropic’s challenge to its Pentagon blacklist designation. The filings argue that the government’s actions may be and potentially punitive, raising concerns about and overreach. The growing institutional backing strengthens Anthropic’s case and signals that the dispute is evolving into a broader legal test of how AI companies can be regulated in national security contexts.
In court filings opposing Anthropic’s lawsuit, U.S. officials argued that the company’s usage restrictions could create operational risks during . The government claimed that Anthropic’s control over and permitted use cases could allow it to limit or disrupt in critical situations, making it an unreliable partner for national security systems. This introduces a new justification beyond contractual disputes, focusing on reliability and control.
The Trump administration filed detailed court arguments defending the Pentagon’s decision to designate Anthropic a . Officials argued the move was based on and concerns rather than free speech, citing Anthropic’s refusal to permit unrestricted ‘lawful use’ of its AI systems by the military. The filing escalates the legal battle, framing the dispute as a precedent-setting case over government authority versus corporate control of AI usage.
OpenAI signed a new agreement to provide its AI models to U.S. via , extending beyond its original Pentagon contract. The deal includes both , marking a significant expansion of OpenAI’s role across federal systems. It follows the removal of from government supply chains and signals growing reliance on OpenAI as a primary AI provider for national security and administrative use cases.
Legal experts and policy analysts began debating whether private should have the authority to restrict how governments deploy their models in national security contexts. The Anthropic-Pentagon dispute raised questions about whether AI vendors can enforce bans on uses such as mass surveillance or autonomous weapons. The case may ultimately determine how much control technology firms retain once their systems are integrated into government infrastructure. The debate is increasingly seen as a precedent-setting test for the governance of military AI.
Palantir CEO Alex Karp acknowledged that the company’s defense analytics platforms continue to rely on Anthropic’s Claude models even after the Pentagon labeled the firm a supply-chain risk. He said the phase-out process will take time because existing systems and software pipelines were built around those models. The admission highlighted the difficulty of removing AI models that have already been embedded in complex defense software stacks. It also demonstrated the broader industry dependence on multiple AI providers rather than a single platform.
Defense contractors and technology suppliers warned that removing Anthropic models from existing systems may disrupt military software used for logistics, data analysis and battlefield planning. Many contractors had integrated models into tools already deployed within defense programs. Industry groups urged the government to provide a longer transition period to avoid operational disruptions while alternative AI providers are integrated.
An internal Department of Defense memo revealed that certain military units may temporarily continue using Anthropic’s AI tools if removing them immediately would disrupt critical operations. Commanders must submit risk-mitigation plans to obtain exemptions while transitioning away from the company’s technology. The guidance reflects how deeply Anthropic’s Claude models are embedded in existing defense systems and contractor software. The Pentagon still requires a full phase-out within roughly 180 days as the blacklist remains in effect.
Multiple congressional committees requested detailed documentation of OpenAI's Pentagon safeguards, including technical specifications for surveillance and autonomous weapons restrictions. Lawmakers expressed concern that contractual red lines lack statutory enforcement mechanisms.
A federal court ruled that Anthropic's lawsuit challenging its supply-chain risk designation can advance to discovery, requiring the Pentagon to disclose internal communications about the blacklisting decision. The ruling strengthens Anthropic's argument that the designation was retaliatory rather than security-based.
Sam Altman faced tough questions from lawmakers in a closed-door Capitol Hill session on OpenAI's deepening Defense Department ties, amid concerns over AI in military applications and transparency. The meeting signals escalating congressional scrutiny on commercial AI labs' defense partnerships. - Sam Altman told lawmakers on United States Capitol that the U.S. must accelerate AI development and infrastructure to stay ahead of China in the global AI race. - He warned that overly strict regulations could slow innovation and argued for “sensible regulation that does not slow us down.” - Altman said AI could unlock major scientific breakthroughs and productivity gains but also carries risks that governments and companies must manage carefully.
Microsoft and dozens of AI researchers filed legal briefs supporting Anthropic’s lawsuit against the Pentagon. They warned the blacklist could disrupt defense contractors using Anthropic technology and argued the government should negotiate safeguards rather than punish companies for imposing ethical limits on AI use.
Anthropic sued the U.S. government after the Pentagon designated the company a 'supply-chain risk' following a dispute over AI safety restrictions. The company argued the decision was retaliatory because it refused to allow its models to be used for mass domestic surveillance or fully autonomous weapons.
OpenAI’s head of robotics, Caitlin Kalinowski, resigned following backlash over the company’s defense agreement with the Pentagon. She warned that surveillance of Americans and lethal autonomous systems deserved deeper debate before deployment.
The U.S. government began drafting new procurement rules requiring AI companies working with federal agencies to allow 'any lawful use' of their models. The guidelines would also require firms to grant the government an irrevocable license to use their systems for legal purposes and disclose regulatory modifications made for other countries.
Sam Altman announced contract modifications explicitly prohibiting use of OpenAI's technology on 'commercially acquired' data, addressing concerns about domestic surveillance that were not covered in the original February 28 agreement. Altman acknowledged he 'shouldn't have rushed' the deal, citing poor optics.
OpenAI CEO Sam Altman conceded the Pentagon deal was rushed, announcing contract amendments to ban domestic surveillance explicitly. The Pentagon affirmed no NSA use of OpenAI tools. A 'Cancel ChatGPT' boycott gained traction over military ties. OpenAI detailed cloud deployment and engineer oversight as safeguards. Altman hosted an X AMA, claiming superior protections versus Anthropic's deal. Calls grew for congressional oversight on AI-defense partnerships.
Anthropic CEO Dario Amodei accused OpenAI of 'straight up lies' in communication about its deal superiority over Anthropic's prior agreement. Tech workers urged Congress and DoD to reverse Anthropic's supply-chain risk label. OpenAI defended its multi-layered safeguards, including cloud deployment and personnel oversight. The feud highlighted diverging AI safety approaches in national security contexts. Tensions underscored ethical debates on AI in defense. TechCrunch and Understanding AI highlight questions on why OpenAI secured red lines rejected for Anthropic.
The White House directed federal agencies to halt use of Anthropic's AI tools after six months, following failed DoD negotiations. This paved the way for OpenAI's rushed Pentagon contract announcement on the same day. The Pentagon escalated by designating Anthropic a supply-chain risk, threatening its U.S. business. OpenAI's Altman later admitted the timing created poor optics amid the Anthropic fallout.
OpenAI announced a contract with the U.S. Department of Defense to deploy its AI models in classified networks, hours after Anthropic refused similar terms over concerns about surveillance and autonomous weapons. CEO Sam Altman highlighted the deal's alignment with safety principles prohibiting mass domestic surveillance and fully autonomous weapons. The agreement emphasizes cloud-only deployment to prevent edge use in weapons systems. OpenAI positioned the partnership as a way to resolve tensions between the DoD and AI companies, urging fair terms for competitors like Anthropic. This move came amid the Pentagon's decision to cut ties with Anthropic because of conflict over safety terms.