The US government has reached an agreement with Anthropic to restore limited access to its Mythos 5 artificial intelligence (AI) model after restricting it earlier over national security concerns, according to The New York Times. Under the deal, the Commerce Department approved the model’s return for select customers after Anthropic agreed to implement additional safeguards and work with the government on future AI releases.

The restrictions were imposed in June after government officials raised concerns that the model could expose vulnerabilities in software code. Anthropic said it is restoring access to approved clients, while negotiations continue over broader access to Mythos 5 and its more advanced Fable 5 model. The agreement follows weeks of discussions between the company and the Trump administration.

According to The New York Times, the deal signals a growing role for the US government in overseeing the release of advanced AI systems. OpenAI has also limited the rollout of its latest model, GPT-5.6 Sol, to government-approved customers, while Meta remains the only major US AI developer that has not agreed to voluntarily submit its models for federal safety review.