Anthropic's Export Control Bind: They Literally Asked for This

Last Friday, the US Government issued an export control directive that bars Anthropic from giving foreign nationals access to Claude Fable or Claude Mythos. Two days earlier, Dario Amodei published a policy essay asking for exactly that power.

The Policy That Came Back to Bite Them

Amodei's "Policy on the AI Exponential" explicitly states: "The government should have the power to block or deter deployment of the model if it is determined, in light of third-party assessment, to present unacceptable risks." The four risks he scopes that power to are cybersecurity, biological weapons, loss of control of AI systems, and automated R&D that accelerates those risks.

Amazon, a frequent government contractor, provided the third-party assessment that Claude Fable and Mythos pose a cybersecurity risk. That gave the Treasury Department, acting through Bessent, the cover it needed. Export control directives are a clear, established power - not the ham-fisted power grab the Department of War attempted earlier this year.

Amodei Asked for This, Letter by Letter

Anthropic spent years pushing for the government to have more control over AI, especially over open source projects and smaller players. They never proposed new protections against political favoritism beyond the ordinary course of law. "Courts," as the author notes, are the protective measures Amodei implicitly trusted. A government lawyer reading his public statements into the record would have a field day.

The irony isn't subtle: Anthropic sold services to Palantir and Amazon, lobbied for heavy regulation, and now finds itself the first target. The leopard ate their face first. What this enables is a clear test case for how the government will use export controls against frontier AI models - and whether Anthropic will finally reconsider its blanket endorsement of regulatory power.

Source: Did Anthropic Ask for This?
Domain: verysane.ai