Altman Calls Pentagon Deal 'Sloppy'
OpenAI CEO Sam Altman admitted the company's recent Pentagon contract “looked opportunistic and sloppy,” a stunning self-critique following the government's exclusion of rival Anthropic. The comment fuels concerns from privacy advocates about concentrating so much government AI reliance on a single provider and the potential risks of a hasty rollout.
The crux of the dispute was Anthropic's refusal to remove safeguards against its AI being used for mass domestic surveillance and in fully autonomous weapons systems. The Pentagon reportedly insisted on contract language allowing for "all lawful uses," which Anthropic argued contained loopholes, especially around the analysis of commercially available bulk data on Americans. In an unprecedented move against a domestic company, the Trump administration designated Anthropic a "supply chain risk," a label typically reserved for foreign adversaries. This designation could legally bar any company that does business with the military from using Anthropic's products, even for non-defense work. Anthropic has stated its intention to challenge the designation in court. Hours after Anthropic was sidelined, OpenAI announced it had secured a deal, initially agreeing to the "all lawful use" standard that its rival had rejected. This move was part of a larger, up to $200 million contract for OpenAI to help the Pentagon prototype frontier AI for applications ranging from administrative tasks to warfighting operations. Following significant backlash from the public and within its own ranks, OpenAI amended the agreement. The new clauses explicitly prohibit the intentional use of its AI for domestic surveillance of U.S. persons, referencing the Fourth Amendment and other laws, and also bar its use by intelligence agencies like the NSA without a separate contract. The controversy has highlighted a policy vacuum for how advanced AI firms should collaborate with governments on defense. The rapid-fire events—from contract negotiations collapsing to a major US AI firm being blacklisted—have spurred a broader debate on the ethics of military AI and the power of private companies to set boundaries on the use of their technology. Despite the ban, reports emerged that the Pentagon used Anthropic's Claude AI during military strikes in Iran just hours after the "supply chain risk" designation was announced, highlighting the model's deep integration into classified workflows. Meanwhile, the public fallout saw a surge in users for Anthropic's Claude, which briefly overtook ChatGPT as the top app in Apple's App Store.
