Donald Trump also signed an executive order this month to create a voluntary national security review system for advanced AI models.
Anthropic has received approval from the US government to let a small group of American cyber security firm’s access for its advanced AI model, Mythos 5.
Earlier, US authorities had blocked access due to national security concerns linked to possible misuse of the system. The company said it will continue talks with the government to expand access to Mythos 5 and may also restore public access to another model, Fable 5, the goal is to gradually reopen access while maintaining safety checks.
In June, US officials ordered Anthropic to cut access to its top AI systems after identifying weaknesses in their safety controls. This sudden move raised concerns in the tech industry and led to criticism that the government was going too far.
Read more: Anthropic restricts new AI models after US government concerns
Anthropic has also faced tension with the Trump administration in recent months. The company refused to support uses such as mass surveillance and autonomous weapons, which led to the Pentagon ending some of its contracts.
At the same time, rival company Open-AI introduced its new model GPT-5.6, but access is being tightly controlled and reviewed on a case by case basis by the government.
President Donald Trump also signed an executive order this month to create a voluntary national security review system for advanced AI models before release. However, the White House has not clearly explained how the system will be enforced or which AI tools will be included.
