HeadlinesBriefing favicon HeadlinesBriefing.com

US blocks Anthropic’s latest AI models

9to5Mac •
×

Anthropic announced it has shut off access to Claude Mythos 5 and Claude Fable 5 following an export control directive from the U.S. Commerce Department. The letter, sent to CEO Dario Amodei late Tuesday, bars any foreign person—inside or outside the United States—from using the two models. All other Anthropic offerings remain online and functional.

The models debuted earlier this week as extensions of the Mythos Preview, a research‑grade system that helped partners locate and patch hundreds of vulnerabilities—Mozilla reported fixing hundreds after using it. Mythos remained limited to Project Glasswing participants, while Fable was released publicly with added guardrails that drew criticism. A reported jailbreak of Mythos sparked the government’s alarm.

The Commerce Department indicated that any export, re‑export, or domestic transfer of the two models now requires a specific license. Officials acted after a competing firm demonstrated a jailbreak that could bypass safety layers, prompting national‑security concerns. The agency previously attempted to block the models during the Trump administration but was unsuccessful.

Anthropic’s response calls the action a misunderstanding, noting the demonstrated jailbreak exposed only minor, already‑known vulnerabilities that other public models can find. The company apologized for the sudden outage and said it will engage with regulators to clarify the issue. For now, developers must pivot to Anthropic’s other models for production workloads.