The Trump administration has introduced fresh guidelines for AI contracts in the civilian sector, mandating companies to allow for the “lawful” utilization of their models. This move comes in the aftermath of a dispute between the Pentagon and AI entity Anthropic. The Pentagon flagged Anthropic as a “supply-chain risk,” prohibiting government contractors from leveraging the AI firm’s technology in projects for the US military. The conflict arose due to disagreements over the level of safeguards requested by Anthropic, which the Defense Department deemed excessive.
According to a report by the Financial Times, a preliminary version of the regulations stipulates that AI entities vying for government contracts must provide the US with an irrevocable license to utilize their systems for all legitimate purposes. The guidance, overseen by the General Services Administration, focuses on civilian contracts and forms part of a broader initiative across the government to enhance the procurement of AI services. This initiative aligns with similar measures being contemplated by the Pentagon for military contracts.
Josh Gruenbaum, the commissioner of the Federal Acquisition Service under GSA, expressed concerns over maintaining a business relationship with Anthropic, emphasizing the need to act in the best interests of the American people and national security. As per Gruenbaum, in compliance with the President’s directive, GSA has terminated Anthropic’s OneGov agreement, thereby restricting their access to the Executive, Legislative, and Judicial branches through GSA’s pre-negotiated contracts.
The White House did not offer an immediate response to Reuters’ requests for comments. The GSA draft mandates that contractors are prohibited from deliberately embedding partisan or ideological biases into the data outputs of AI systems. Additionally, companies must disclose whether their models have undergone alterations to conform with any non-US federal government or commercial compliance frameworks.