White House Proposes Tougher Rules for AI Contracts
The White House has introduced proposals to tighten rules for civilian AI contracts, aiming to improve security and oversight. The move follows the recent Pentagon clash with AI firm Anthropic and signals a new government-wide push to control how powerful AI models are used in sensitive domains.
The new draft rules for civilian AI contracts come from the General Services Administration (GSA) and mandate that tech firms grant the U.S. government an irrevocable license to use their AI systems for "any lawful purpose." This move is seen as part of a broader government effort to standardize and tighten the procurement process for AI services across all civilian agencies. This policy mirrors similar measures being weighed by the Pentagon for military contracts, indicating a new, more assertive government-wide stance on the use of AI. The push for "any lawful use" grew directly out of a recent dispute with AI firm Anthropic, which had sought to include contractual limitations on the use of its technology by the Department of Defense. Anthropic expressed specific concerns about its AI, named Claude, being used for domestic mass surveillance or to direct fully autonomous weapons systems without human oversight. The Pentagon, however, insisted on the option to use the technology for all lawful purposes, arguing that existing laws and internal policies already provide sufficient safeguards. The standoff led to the Pentagon canceling a significant contract with Anthropic and taking the unusual step of designating the American company a "supply-chain risk," a label typically reserved for firms with ties to foreign adversaries. In response to the dispute, President Trump ordered all federal agencies to cease using Anthropic's technology. The draft GSA guidance also includes provisions aimed at ensuring political neutrality. Contractors would be required to prevent their AI systems from encoding "partisan or ideological judgments" and to create a "neutral, non-partisan tool that does not manipulate responses in favour of ideological dogmas such as diversity, equity, inclusion." This reflects a broader policy initiative from the Trump administration, which has previously issued executive orders aimed at what it terms "woke AI" in federal procurement. Companies would also have to disclose if their models have been modified to comply with any foreign government or commercial regulations.
