Two weeks ago, Anthropic’s most potent cybersecurity models, Mythos 5 and Fable 5, were effectively pulled from the grid. The ban, which barred even non-American employees from accessing the software, sent a chill through the AI sector. Now, the Trump administration is reversing course.

Commerce Secretary Howard Lutnick confirmed on Friday that the government has authorized the redeployment of Mythos 5 to more than 100 U.S. agencies and private companies. The directive, outlined in a letter to Anthropic Chief Compute Officer Tom Brown, explicitly permits these organizations to grant access to their non-American staff—a significant departure from the strict national-origin restrictions imposed earlier this month.

Why the Pivot Matters

The decision signals a shift in how the administration views the intersection of national security and AI development. By allowing global teams at critical infrastructure firms to utilize Mythos 5, the White House is acknowledging that the defense of U.S. digital assets is a borderless endeavor. The model, designed specifically to identify and patch vulnerabilities in complex codebases, is considered a vital tool for firms operating power grids, financial networks, and telecommunications.

However, the administration’s approval comes with strings attached. The authorization is limited to "trusted partners" that have satisfied new, rigorous safety protocols. While the government has yet to release the full list of these 100-plus entities, the directive suggests that the Commerce Department is moving toward a tiered access model rather than a blanket ban.

The Missing Piece: Fable 5

While Mythos 5 is returning to the field, Fable 5 remains sidelined. The administration’s directive was notably silent on the status of the second model, which was originally marketed as a more secure, guardrail-heavy alternative to Mythos. Both models were pulled from the market after security researchers demonstrated that their safety protections could be bypassed with relative ease.

Anthropic has been scrambling to regain the government's trust. In a post on X, the company confirmed it is working to restore access for the approved organizations while continuing negotiations to bring Fable 5 back into the fold. For now, the company’s strongest cybersecurity asset is back, but the regulatory scrutiny remains intense.

What This Means for Users

For the organizations involved, the return of Mythos 5 is a reprieve from a two-week operational bottleneck. Security teams that rely on the model to automate threat detection can now resume their workflows, albeit under the watchful eye of the Commerce Department.

However, the broader developer community remains in limbo. The restriction on Fable 5—the version intended for wider, general use—suggests that the administration is not yet convinced that Anthropic’s safety architecture is robust enough for public consumption.

Key Takeaways

  • The Trump administration has authorized the use of Mythos 5 for over 100 U.S. agencies and critical infrastructure firms, including their international staff.
  • The directive specifically excludes Fable 5, which remains under a government-imposed hiatus following reports of bypassed safety guardrails.
  • Access is now contingent on "appropriate safeguards," marking a shift toward a restricted, vetted deployment model for high-stakes AI tools.

The next major hurdle for Anthropic is the upcoming quarterly review of its safety protocols, scheduled for late next month. Whether Fable 5 is cleared for general release will depend on whether the company can prove to the Commerce Department that its latest patches can withstand the same adversarial testing that led to the initial ban.