OpenAI’s Sam Altman announces Pentagon deal with ‘technical safeguards’

On Friday, Sam Altman, the CEO of OpenAI, revealed that his organization has finalized a deal permitting the Department of Defense (DoD) to utilize its AI models within the military's classified network. This announcement follows a notable conflict between the Pentagon and OpenAI's competitor, Anthropic, which had previously resisted similar requests. The Pentagon had been urging AI firms, including Anthropic, to allow their technologies to be used for 'all lawful purposes.' However, Anthropic raised concerns about the potential implications of their models in contexts of mass domestic surveillance and fully autonomous weaponry. In a detailed statement, Anthropic's CEO Dario Amodei emphasized that while they did not object to specific military operations, they believed that AI could sometimes compromise democratic values. In a show of solidarity, over 60 employees from OpenAI, alongside 300 from Google, signed an open letter advocating for Anthropic’s stance. Following the breakdown of discussions between the Pentagon and Anthropic, former President Donald Trump criticized the company on social media, labeling it as a group of 'Leftwing nut jobs' and calling for federal agencies to cease using its products within a six-month timeframe. In a retaliatory move, Secretary of Defense Pete Hegseth accused Anthropic of attempting to "seize veto power over the operational decisions of the United States military" and designated the company as a supply-chain risk, effectively barring any contractors working with the military from engaging commercially with Anthropic. Despite these developments, Anthropic stated it had not yet received any formal communication from the Pentagon or the White House regarding the status of their negotiations, and indicated a willingness to contest the supply-chain risk designation in court. In a surprising twist, Altman announced that OpenAI's defense contract includes specific safeguards addressing the same contentious issues that had been pivotal in Anthropic's discussions. He highlighted two fundamental safety principles: a ban on domestic mass surveillance and the necessity for human oversight in the use of force, particularly concerning autonomous weapon systems. According to Altman, the DoD has acknowledged these principles, which are now embedded in their agreement. Altman further stated that OpenAI would implement technical safeguards to ensure that its models function correctly, and that the company would assign engineers to work closely with the Pentagon to oversee the models' safety. He expressed a desire for the DoD to extend similar terms to all AI developers, advocating for a shift from legal confrontations to constructive agreements. In an internal meeting, Altman informed OpenAI staff that the government would permit the company to create its own safety mechanisms to prevent misuse of its technology, asserting that if a model declined to perform a task, the government would not compel OpenAI to override that decision. This announcement coincided with alarming news of escalating military actions involving the U.S. and Israel against Iran, alongside Trump’s calls for regime change in the country.