U.S. Government Lifts Its Block on Anthropic's AI Models — With Strings Attached

Key points

• The U.S. Commerce Department banned foreign nationals from using two Anthropic AI models on June 12, forcing the company to pull both products for all users worldwide.
• Amazon researchers found a method to bypass safety filters in Anthropic's Claude Fable 5 model that could allow the AI to find and exploit software flaws in real systems.
• As of Tuesday, Claude Fable 5 is publicly available again; the more powerful Mythos 5 remains limited to a small group of U.S. organisations approved by the federal government.
• Rival AI company OpenAI also restricted its new GPT-5.6 Sol model to government-approved customers following a similar request from the Trump administration.
• President Trump signed an executive order last month allowing the government up to 30 days to review advanced AI systems for national security risks before public release.

Anthropic, the San Francisco company behind the Claude family of AI chatbots, spent several weeks shut out of its own product launch. Now it's getting most of that access back — though not all of it, and not without conditions.

The trouble started June 12, when the U.S. Commerce Department blocked foreign nationals from using two of Anthropic's newest AI models. Anthropic says that order forced the company to pull both products entirely, for every user everywhere, just days after it had unveiled them.

The reason: researchers at Amazon — which provides the computing infrastructure Anthropic runs on — had discovered that one of the models could be manipulated in a way that bypassed its built-in safety guardrails, the rules designed to stop the AI from doing harmful things. Once those guardrails were gone, the model could find weaknesses in software systems and potentially help someone exploit them.

How dangerous is an AI that can find software flaws?

Very, if it falls into the wrong hands. Finding holes in software — what security professionals call vulnerabilities — normally requires significant expertise and time. An AI model that does it quickly and automatically hands that capability to anyone with access, including criminals or foreign spy agencies.

Anthropic had already flagged this concern itself earlier this year, warning that its more powerful model, Mythos 5, was unusually good at spotting those flaws. The Amazon researchers' report appears to have pushed the government from watching closely to acting fast.

The restrictions tracked loosely with an executive order President Trump signed last month. That order created a framework letting the federal government review the most advanced AI systems for national security risks for up to 30 days before they go public. Participation is described as voluntary — but Anthropic and its chief rival OpenAI, maker of ChatGPT, both complied.

OpenAI said Friday its new model, GPT-5.6 Sol, will stay limited to government-approved customers for now.

For ordinary people who use Claude for writing, research, or coding help, the practical upshot is straightforward: the standard version is back. If you were locked out, you should have access again. The advanced Mythos 5 model remains restricted, and there is no public timeline for when that changes.

If you work at an organisation that was evaluating either model for internal use, check with your vendor contact about current access tiers — and think carefully about what data you feed any AI system while these questions about safety controls are still being worked out.