Federal oversight eases for Anthropic as Cybersecurity model returns to service

One hundred organizations, including members of the Fortune 500, regained access on Friday to a tier of intelligence previously withheld under national security directives. The government’s notification to Anthropic allows for the redeployment of Mythos 5, a model designed specifically for cybersecurity defense, to entities deemed trusted partners. This move follows a June 12 export control order that had effectively silenced the company’s most capable systems for all users.

The tension between rapid commercial deployment and state-level caution is tightening. While Anthropic reopens access to its cybersecurity model, OpenAI has concurrently confirmed a delay in the public launch of its GPT-5.6 system at the government’s request. Both firms are navigating a newly established federal framework that requires developers of covered frontier models to provide a 30-day review period before public release, a strategy intended to assess risks in software vulnerability identification and potential military misuse.

Critique of the current vetting process centers on its opacity. Industry leaders have noted that the criteria for selecting the authorized companies remain private, raising questions about the influence of government discretion within the technology market. Current restrictions remain in place for Anthropic’s Fable 5 model, which shares the same underlying architecture as Mythos but lacks the specific protective constraints required for wider use. For now, the administration’s directive functions as an interim compromise, attempting to balance the speed of American development against the hazards of sophisticated, automated cyberattacks.