While everyone was watching the Pentagon fight Anthropic in open court, another federal agency was quietly rewriting the rules to make sure that fight never happens again. The General Services Administration just proposed new AI procurement guidelines. The EFF, CDT, Protect Democracy, and EPIC filed joint comments this week calling them dangerous. They are not overreacting.

The most alarming provision: any AI company that sells to the federal government would be required to license its technology for all lawful purposes. That sounds reasonable until you think about what it means. Mass surveillance is lawful. Predictive policing is lawful. Using AI to process asylum claims at scale is lawful. An AI company could build a medical diagnosis tool and the government could repurpose it for border enforcement.

The context matters. Anthropic refused to let the Pentagon use Claude for military targeting and surveillance. The Pentagon responded by trying to blacklist them from all government contracts. A federal judge blocked it. But these new GSA rules would make that kind of refusal contractually impossible in the future. If you sell to the government, you surrender control over how your AI gets used.

The proposed rules also include provisions requiring AI tools to be ideologically neutral, which the nonprofits warn could be weaponized against safety features. Content moderation? That could be framed as ideological bias. Refusing to generate harmful content? Potentially non-neutral. The vagueness is the point.

If adopted, these rules would become standard components of every federal contract. That is not a small market. The federal government is the single largest buyer of technology services in the world. Any AI company that wants government revenue would have to accept these terms. And any company that refuses gets locked out entirely.

The strategy is elegant in its brutality. You do not need to pass a law. You do not need Congressional approval. You just rewrite procurement rules and let the market do the enforcement. Companies that comply get billions in contracts. Companies that resist get nothing. The Anthropic situation resolved through raw economic pressure instead of legal precedent.

This is happening while California pushes in the exact opposite direction with its own AI procurement standards, demanding companies explain their policies on bias, civil rights, and content distribution. Two competing visions of how government should buy AI. One demands safety transparency. The other demands unconditional compliance.

Every AI company building tools for enterprise and government clients should be reading the GSA proposal right now. Because the next Anthropic will not get a court hearing. The contract will already have been signed.