The U.S. government is preparing tougher rules for artificial intelligence companies seeking federal contracts, a move that could reshape how AI firms work with Washington. The proposed guidelines come amid a growing dispute between the Pentagon and AI company Anthropic, highlighting the tensions between national security demands and tech industry safeguards.
According to a Financial Times report, the Trump administration has drafted new requirements that would force AI companies to grant the U.S. government broad rights over how their models can be used.
New Rules for AI Contracts
The draft guidelines reportedly state that companies supplying AI technology to the U.S. government must allow “any lawful” use of their systems.
In practice, this means AI developers would have to provide the government with an irrevocable license to deploy their models for all legal purposes under federal law.
The policy is being developed by the General Services Administration (GSA) and would initially apply to civilian government contracts, though similar requirements are reportedly being considered for military agreements as well.
The move is part of a broader effort by Washington to strengthen procurement rules for AI services as the technology becomes increasingly central to government operations.
Pentagon Labels Anthropic a Supply Chain Risk
The policy debate intensified after the U.S. Department of Defense designated Anthropic as a “supply-chain risk.”
The decision means government contractors are now barred from using Anthropic’s technology for military-related work.
The restriction follows a months-long disagreement between the Pentagon and the company. Anthropic had reportedly insisted on strict safeguards governing how its AI systems could be used by the Defense Department.
U.S. officials argued that those limitations were too restrictive and could hinder military operations.
Government Ends Anthropic Procurement Deal
The dispute has already had immediate consequences.
According to Josh Gruenbaum, commissioner of the Federal Acquisition Service within the GSA, the government has terminated Anthropic’s participation in the OneGov procurement program.
That program previously allowed the company’s AI services to be purchased by agencies across the executive, legislative, and judicial branches through pre-negotiated government contracts.
Gruenbaum said continuing a business relationship with the company would be irresponsible given the concerns raised by federal authorities.
New Requirements for AI Developers
The draft policy also introduces additional rules aimed at preventing political or ideological bias in government-used AI systems.
Under the proposal, contractors would be required to ensure that their AI models do not intentionally encode partisan or ideological viewpoints into their outputs.
Companies would also need to disclose whether their systems have been modified to comply with regulations from foreign governments or non-U.S. regulatory frameworks.
These transparency requirements reflect Washington’s growing focus on ensuring that AI technologies used by federal agencies align with national security and policy priorities.
Rising Tensions in the AI Policy Debate
The clash between the Pentagon and Anthropic highlights a larger issue confronting the AI industry: balancing safety safeguards with government demands for flexibility.
Many AI companies have introduced internal restrictions designed to prevent misuse of their models. However, governments — especially defense agencies — often want fewer limitations so they can apply AI tools in a wider range of scenarios.
As AI rapidly becomes a critical part of national security, procurement policy is emerging as a key battleground between tech companies and government regulators.
Final Words
With the U.S. government tightening its approach to AI procurement, companies seeking federal contracts may soon face stricter oversight and broader licensing requirements. The proposed rules signal that Washington intends to maintain strong control over how artificial intelligence is deployed within federal agencies.
At the same time, the dispute with Anthropic shows how complex the relationship between AI developers and governments has become — especially when issues of security, ethics, and technological power collide.
