Room seven-b · inside the rulebook fight
The Pentagon ended a $200 million contract with Anthropic and ordered other military contractors to stop using its products — after Anthropic refused to drop restrictions on using Claude for mass domestic surveillance.
Dr. Matthew Guariglia, Electronic Frontier Foundation, testifying to the House Homeland Security Subcommittee, June 4, 2026
"AI also has a track record of getting things wrong — from false citations on legal briefs to a major AI mistake that sent DHS recruits to the field without proper training." Asked how to rein in AI, his answer reframed the question entirely: "At this level the question is not how do we rein in AI, it's how do we rein in the agencies that would unleash AI on the American public."
U.S. District Court, Northern District of California, March 24, 2026
Granted Anthropic's motion for a preliminary injunction, finding the government's actions "were not designed to protect national security, but rather to punish Anthropic," calling it "classic illegal First Amendment retaliation."
EFF's own position, stated directly, that complicates a simple reading
EFF is not arguing for more government control over AI companies in general. Its written testimony states plainly: "It is essential to preserve the rights of companies to conduct frontier AI research, freely develop multi-use models for the public, and publicly release new models without the need to ask the government for permission to do so." The group that defended Anthropic here is the same group that would resist a future government license requirement on any AI lab, including ones with a worse safety record.
Anthropic refused the Pentagon and a court agreed it was retaliation. The same week, the same government can still block Anthropic's own model exports by executive order. Is that the same government acting two different ways, or two different parts of government that don't agree with each other either?