Altman Calls Pentagon Deal 'Sloppy'
OpenAI CEO Sam Altman has acknowledged the company's recent Pentagon deal looked “opportunistic and sloppy.” The admission comes amid new reports that OpenAI “caved” to military demands for broad access and surveillance capabilities. The company has also struck a new deal to deploy its AI on classified networks for the Defense Department's DOW division.
The controversial deal unfolded rapidly after the Pentagon's previous AI partner, Anthropic, refused to alter its ethical guidelines. Anthropic insisted on explicit contractual bans on using its AI for mass domestic surveillance or for autonomous weapons that can kill without human oversight. This led President Donald Trump to label the company "leftwing nut jobs" and order all federal agencies to cease using its technology, designating Anthropic a "supply chain risk." Hours after the public fallout with Anthropic, OpenAI announced it had secured a deal to deploy its models on the Pentagon's classified networks. The agreement is part of a broader push by what the Trump administration refers to as the Department of War (DOW) to accelerate its AI dominance, with the Pentagon's Chief Digital and AI Office leading the charge. OpenAI is set to "develop prototype frontier AI" under a contract worth up to $200 million. Initial criticism centered on contract language that permitted the Pentagon to use OpenAI's technology for "all lawful purposes." This was a key point of contention that Anthropic had rejected, fearing it created loopholes for surveillance and autonomous weapons use. OpenAI's agreement, in contrast, was seen as much "softer" and more accommodating to the military's demands. The backlash was immediate and widespread, with reports of a 295% increase in ChatGPT uninstalls as users flocked to rival AI Claude, which temporarily topped Apple's App Store charts. Internally, nearly 100 OpenAI employees and close to 800 Google employees signed an open letter warning against a race to the bottom and urging their leaders to refuse demands for surveillance and autonomous killing capabilities. In response to the outcry, OpenAI amended its agreement with the Pentagon. The new language explicitly bars the use of its technology for the intentional domestic surveillance of U.S. citizens and prohibits its use by intelligence agencies like the National Security Agency (NSA) without a separate agreement. Despite the changes, OpenAI's deployment will be cloud-only, a measure the company claims prevents its models from being used on "edge devices" such as autonomous drones. The company will also have cleared engineers on-site to help oversee the government's use of the technology.
