June 12, 2026, in response to an urgent announcement jointly issued by the US Department of Commerce and the National Security Council, Anthropic removed its newly released Claude 5 Fable and Claude 5 Mythos 5 models from public availability pending further clarification of US government requirements. The announcement explicitly states that non‑US citizens or non‑US lawful permanent residents (including foreign nationals holding visas such as H‑1B or L‑1) are prohibited from directly accessing or using the models' full capabilities. The move quickly sent shockwaves through the global AI community and is seen as another major escalation in US export controls on AI technology.

Anthropic issued a statement later the same day saying it will strictly comply with US government requirements and has adjusted its API and Claude.ai platform accordingly. At present, the highest‑performance versions of Claude 5 Fable and Mythos 5 are limited to users located in the United States, US‑based companies, and approved allied partners.

Details of the restrictions

According to the US Department of Commerce notice, the main measures include:

Prohibition on foreign nationals directly invoking the full inference capabilities of the two models (including highest‑performance modes and long‑context features).
Enterprise access must be requested through a US entity and is subject to stringent end‑user checks.
Open‑source weight releases are also restricted; overseas developers must obtain special authorization to download and use them.
Violations may result in civil and criminal penalties, including heavy fines and placement on export control blacklists.

Anthropic said it has simultaneously disabled advanced access channels for overseas users and launched two downgraded versions for international use: “Claude 5 Fable‑Lite” and “Opus 4.8 Safe.”

Policy background and strategic considerations

The immediate context for the restrictions is ongoing US concern about the spread of advanced AI capabilities. Claude 5 Fable and Mythos 5 are regarded as among the world’s most capable long‑horizon agentic models, with strong autonomous planning, multi‑step reasoning, and complex task execution. US officials worry that if such capabilities are obtained by potential adversaries, they could be repurposed for military, intelligence, or critical‑infrastructure attacks.

US Commerce officials have emphasized in internal briefings that “the Claude 5 series marks a critical juncture in AI’s shift from tools to agents; its strategic importance is comparable to that of nuclear technology non‑proliferation.” The policy follows prior restrictions on companies such as Huawei and TikTok and reflects a tightening of the US “AI national security strategy.”

Strong global reactions

China’s tech sector: Several companies and research institutions expressed strong dissatisfaction, calling the move a “technology barrier” and “digital hegemony.” Alibaba, Baidu and MiniMax quickly issued statements saying they will increase investment in domestic large‑model development. Some developers said they will shift focus to open‑source models and domestic alternatives.

European and Asian allies: The European Commission and Japan’s Ministry of Economy, Trade and Industry both expressed regret, warning that the US action could undermine international AI cooperation. Some countries have begun discussions with Anthropic on an “allies exemption” mechanism.

Industry voices: OpenAI CEO Sam Altman said he “understands US national security concerns but hopes to preserve openness of technology”; Google DeepMind said it will strengthen technical sharing