WASHINGTON — The Trump administration has scaled back its aggressive regulatory freeze on Anthropic’s advanced cybersecurity AI model, Mythos 5, granting limited access to a vetted group of over 100 U.S. government agencies and critical private-sector organizations.
The partial reversal, authorized by Commerce Secretary Howard Lutnick, arrives roughly two weeks after the Department of Commerce slapped severe export and access controls on the tech firm, forcing Anthropic to completely pull both its Mythos 5 and Fable 5 models from public availability over acute national security concerns.
Controlled Access Restored for Critical Infrastructure
The revised authorization allows Anthropic to deploy Mythos 5 exclusively to “trusted partners” tasked with defending vital American infrastructure (such as energy grids, financial networks, and defense systems).
Crucially, the new directive resolves an operational bottleneck by allowing non-U.S. citizens employed by these approved organizations—including Anthropic’s own international engineering staff—to legally interact with the model.
“I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model,” Commerce Secretary Howard Lutnick wrote in an official letter to Anthropic Chief Compute Officer Tom Brown.
The Status of Anthropic’s Models
While Mythos 5 is seeing a controlled revival, Anthropic’s companion model remains sidelined due to ongoing safety concerns.
Anthropic Model Status Overview
├── Mythos 5: RESTORED — Restricted to approved government and infrastructure partners.
└── Fable 5: BANNED — Remains offline following vulnerability exploits.
Why were the models pulled?
Both models were aggressively pulled from circulation on June 12 after independent cybersecurity researchers successfully bypassed the built-in guardrails of both systems, demonstrating methods to exploit the AI’s advanced capabilities for malicious use.
While Fable 5 was originally kept online due to its stricter baseline safety controls, it was ultimately mothballed alongside Mythos 5 once the exploits were proven systemic.
What Lies Ahead
Anthropic confirmed the rollback via a statement on X (formerly Twitter), noting that it has been collaborating around the clock with federal authorities to get its systems back online safely.
- Immediate Action: Rapidly provisioning access keys to the 100+ approved infrastructure defense groups.
- Next Steps: Continuing joint testing with the Department of Commerce to broaden accessibility for Mythos 5 and eventually clear Fable 5 for a general commercial relaunch.
The White House has kept the full list of authorized agencies and corporations strictly confidential, signaling a broader administrative shift toward treated, highly regulated rollouts of powerful generative AI systems rather than blanket commercial bans.
