Anthropic has begun restoring access to its most advanced AI models following a temporary suspension earlier this month due to U.S. government security concerns. According to a report by Axios on June 27, 2026, the company has officially restored the Mythos 5 model for a limited group of trusted users. Mythos 5, which specializes in cybersecurity, was originally designed with additional guardrails to mitigate risks related to biological threats and cyberattacks. Unlike the company's broader releases, Mythos 5 remained restricted from public access.

The return of Fable 5, Anthropic's flagship model for coding and deep reasoning, remains pending as negotiations with the U.S. administration continue. Access to Fable 5 was abruptly pulled just days after its launch, leaving developers who had integrated the system into their workflows without service. Commerce Secretary Howard Lutnick indicated that Anthropic has made substantial progress in addressing government safety concerns. Final approval for Fable 5 currently depends on evaluations from several federal agencies, including the Pentagon and the National Security Agency, with a potential restoration expected as early as the week of June 29, 2026.

The phased restoration strategy reflects a shifting regulatory environment for frontier AI systems. By prioritizing the return of the more restricted Mythos 5, regulators appear to be demonstrating a preference for models equipped with rigorous safety guardrails. This trend extends beyond Anthropic, as seen with OpenAI's recent release of its GPT-5.6 preview, which similarly limits access to select partners during government review processes. These developments indicate that launching state-of-the-art AI systems now requires navigating complex regulatory milestones alongside traditional engineering benchmarks.

Anthropic has begun restoring access to its most advanced AI models following a temporary suspension earlier this month due to U.S. government security concerns. According to a report by Axios on June 27, 2026, the company has officially restored the Mythos 5 model for a limited group of trusted users. Mythos 5, which specializes in cybersecurity, was originally designed with additional guardrails to mitigate risks related to biological threats and cyberattacks. Unlike the company's broader releases, Mythos 5 remained restricted from public access.

The return of Fable 5, Anthropic's flagship model for coding and deep reasoning, remains pending as negotiations with the U.S. administration continue. Access to Fable 5 was abruptly pulled just days after its launch, leaving developers who had integrated the system into their workflows without service. Commerce Secretary Howard Lutnick indicated that Anthropic has made substantial progress in addressing government safety concerns. Final approval for Fable 5 currently depends on evaluations from several federal agencies, including the Pentagon and the National Security Agency, with a potential restoration expected as early as the week of June 29, 2026.

The phased restoration strategy reflects a shifting regulatory environment for frontier AI systems. By prioritizing the return of the more restricted Mythos 5, regulators appear to be demonstrating a preference for models equipped with rigorous safety guardrails. This trend extends beyond Anthropic, as seen with OpenAI's recent release of its GPT-5.6 preview, which similarly limits access to select partners during government review processes. These developments indicate that launching state-of-the-art AI systems now requires navigating complex regulatory milestones alongside traditional engineering benchmarks.