Anthropic just made government ID the price of admission for its best AI model. Your vendor strategy has a new dependency.

The most capable AI model ever released was killed by a government email.

On June 12, the US Commerce Department gave Anthropic 90 minutes to disable Fable 5 and Mythos 5 globally.

Not for foreign users. For everyone.

The trigger: Amazon researchers found that asking Fable 5 to "fix this code" instead of "review this code" bypassed safety guardrails. Amazon CEO Andy Jassy briefed Treasury Secretary Bessent. The government called it a national security risk.

Anthropic complied. Then published a statement saying the jailbreak was narrow, non-universal, and identical to capabilities in GPT-5.5 — which was not banned.

Here is what happened next.

Anthropic updated its privacy policy. Effective July 8, the company can now collect government-issued ID, facial geometry templates, and biometric data from Free, Pro, and Max users.

The verification runs through Persona — a Peter Thiel-backed identity vendor.

The logic is straightforward: if Anthropic can prove only verified US persons access Fable 5, it potentially satisfies the export control directive without waiting for the government to formally lift it.

Your vendor strategy now has three new dependencies.

First, model availability is no longer a technical question. It is a regulatory one. The same model your engineers built workflows around can be disabled by a single Commerce Department letter.

Second, identity verification is becoming the gate for frontier AI access. Anthropic is first. OpenAI and Google are moving the same direction. Your zero-retention data policies may need rewriting.

Third, your AI vendor's relationship with the US government is now a procurement risk. Cohere is already fielding inbound calls from enterprises reassessing their Anthropic dependency.

Audit your Claude dependencies today. Identify which workflows depend on Fable 5 class capabilities. Build tested fallback paths to Opus 4.8 or alternative vendors. Review your data processing agreements before July 8.

The era of "plug in an API key and forget about it" is over. Your AI stack now has political risk built in.

SOURCE: https://www.anthropic.com/news/fable-mythos-access
VERIFIED: Anthropic official statement (June 12), Bloomberg Lutnick letter publication (June 16), Axios Trump interview (June 19), explainx.ai ID verification report (June 21)
SIGNAL: First government-ordered frontier model recall. ID verification as access gate sets precedent for every AI vendor. Enterprise continuity plans must account for regulatory-driven model withdrawal.