Corcos’ banishment from AI policy followed multiple points of friction with the White House, including his role in greenlighting Anthropic’s Fable model days before Amazon reported a jailbreak, and his more recent unsanctioned outreach to technology executives.
The person familiar with the matter said Corcos convened conversations with executives at OpenAI and Anthropic, as well as members of their teams, without giving the White House full visibility.
During those conversations, the person said, Corcos portrayed himself as an industry ally and warned that other government officials, reacting to fears over the latest AI models, were attempting to impose a burdensome licensing regime that would hamper innovation.
Those discussions occurred at a delicate moment for top White House staff. AI companies were rolling out advanced new models, and the administration was struggling to coalesce around a coherent regulatory strategy. Some inside the White House favored more government intervention in assessing which models can be released and when, while others advocated for a voluntary framework to test model safety.
“He was telling industry that the administration was trying to screw them,” the person said. Corcos, the person added, presented himself as “the only one who was for a free and open process.”
A Treasury spokesperson said in a statement, “As Chief Information Officer, Sam Corcos has been a trusted advisor and has played an important role in modernizing the Treasury Department’s technology and operations. He remains involved in providing policy guidance on issues within his area of expertise.” The spokesperson confirmed that Corcos continues to lead a dedicated team with “significant resources focused on replacing antiquated systems, streamlining operations, and delivering substantial savings for American taxpayers.”
The move is the latest example of the White House struggling to coalesce around an AI policy as it grapples with how to regulate increasingly ubiquitous and dangerous artificial intelligence models. A former Anthropic employee’s decision to quit this week and warn on X that AI “could kill us all by the end of the decade” has further fueled concern in Congress, and Democrats are laying the groundwork for a select committee on artificial intelligence should they control the House in January.