USA maintains restrictions on foreign access to Anthropic's Mythos 5 model

The Trump administration continues to restrict access by foreign governments, companies, and financial institutions to the Mythos 5 AI model from Anthropic PBC, even after lifting export restrictions on other recent models from the company. This is reported by Bloomberg.

Who remains in limbo

Access to Mythos 5 is currently permitted only to a limited circle of organizations within the USA. Foreign governments, companies, and financial institutions that expected to gain access to the model along with the easing of Anthropic's general export rules found themselves in a state of uncertainty — Bloomberg describes their position with the word "limbo."

• The restriction concerns specifically the Mythos 5 model, not the entire Anthropic product line
• Export restrictions on other recent models from the company have already been lifted by the Trump administration
• Access to Mythos 5 remains only with a narrow circle of organizations in the USA
• Among affected parties are foreign governments, commercial companies, and financial institutions

Why the USA selectively restricts AI models

Anthropic PBC is one of the leading American AI labs, founded in 2021 by Dario and Daniela Amodei, former OpenAI executives; the company develops the Claude model family and is considered one of the key players on the AI frontier alongside OpenAI, Google DeepMind, and xAI. Washington has been using export control as a tool for technological containment for several years: first it was restrictions on supplying advanced Nvidia chips to China, then — multilevel rules for access to the AI models themselves depending on the country of the recipient. Throughout 2025, the Trump administration has repeatedly revised the rules inherited from Biden's team, easing one export while maintaining another.

The separate mention of financial institutions is noteworthy: banks and investment funds around the world are increasingly using AI models for risk analysis, compliance, and algorithmic trading, so restricting access specifically to Mythos 5 directly affects their technology plans, even if other Anthropic products are already open to them.

How model control differs from chip control

Export control over ready-made AI models works differently from control over Nvidia chips: if you can restrict the supply of hardware at the customs and manufacturer level, restricting access to the model itself means controlling the distribution of weights, API keys, and cloud access to inference. This is precisely why decisions about which models can be exported and which cannot are often made on a case-by-case basis and may appear inconsistent from the outside: authorities assess not only the technical power of the model, but also what capabilities it opens up — for example, the ability to accelerate biological, cyber, or military development.

What this means

The case-by-case, rather than package, approach to AI model export restrictions means that general easing of policy does not guarantee foreign partners predictable access to specific developments — companies and governments outside the USA will have to orient themselves by each Anthropic model separately, rather than by policy as a whole.