Artificial IntelligenceCybersecurityNewswireTechnologyWhat's Buzzing

US approves Anthropic’s Mythos 5 access for trusted cyber defenders

▼ Summary

– Commerce Secretary Howard Lutnick cleared Anthropic to redeploy Mythos 5, its most powerful cybersecurity model, to a select group of trusted cyber defenders.
– The restriction on Fable 5, the public-facing version of the model, remains unchanged, and it is still unavailable to general users.
– The clearance partially resolves a confrontation that began two weeks ago when the government forced Anthropic to disable both models over security concerns.
– Anthropic is continuing talks with the administration to expand Mythos 5 access and restore Fable 5, with discussions expected to continue over the weekend.
– The partial restoration aligns with a pattern where the Trump administration manages access to frontier AI models through government-approved lists, as seen with OpenAI’s similar limitation.

The U.S. government has authorized Anthropic to reinstate access to Mythos 5, its most advanced cybersecurity model, for a limited group of vetted partners. In a letter dated Friday and reviewed by Bloomberg, Commerce Secretary Howard Lutnick informed Anthropic co-founder Tom Brown that the company’s work to resolve security issues had “yielded significant progress,” allowing the model to be shared with “certain trusted partners.” The letter does not indicate any shift in policy regarding Fable 5, the public-facing version of the same model, which remains restricted.

This decision partially resolves a standoff that erupted two weeks ago when the government invoked export controls to compel Anthropic to disable both Mythos 5 and Fable 5, citing concerns that security guardrails could be bypassed. Anthropic cut off global access to both models because it could not instantly distinguish between domestic and foreign users.

Anthropic stated that Mythos 5 “can be redeployed to a small group of cyber defenders and infrastructure providers,” and the company is working to restore their access as quickly as possible. The firm expressed satisfaction with “this progress” and said it continues collaborating with the government to broaden Mythos 5 access and eventually make Fable 5 available to the general public again.

It remains unclear exactly which steps Anthropic took to address the government’s jailbreaking concerns. Discussions between the company and the administration are expected to continue over the weekend, with the goal of restoring Fable 5 access soon, according to a source familiar with the matter. Anthropic is also working with officials to develop a policy framework for handling similar situations in the future.

The negotiations involved senior Anthropic executives, including co-founder Tom Brown, who met with Lutnick and other administration officials in recent days, Bloomberg reported. CEO Dario Amodei stayed largely out of the talks, a move that insiders said helped reduce tensions between the two sides. This approach contrasts with Amodei’s direct engagement with the White House earlier this year, when he met with Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent over the broader Mythos dispute.

The partial restoration comes on the same day that rival OpenAI limited the release of its own flagship model under similar government pressure, rolling out a preview to partners approved on a customer-by-customer basis by the administration. Together, these moves suggest the Trump administration is establishing a pattern where frontier AI models reach the market only through government-managed access lists, at least during an initial period.

Before the shutdown, roughly 200 firms had access to Mythos through Project Glasswing, Anthropic’s invite-only cybersecurity program, including Apple, Google, Cisco, Nvidia, Microsoft, and JPMorgan Chase. The restriction on Fable 5 means that hundreds of millions of regular users who previously had access to Anthropic’s most capable public model still cannot use it, weeks after the company confidentially filed for an IPO at a valuation exceeding $900 billion.

(Source: The Next Web)

Topics

ai regulation 95% mythos 5 redeployment 92% fable 5 restrictions 88% government-industry negotiations 85% national security concerns 82% trusted partner access 80% anthropic company strategy 78% howard lutnick role 75% tom brown leadership 72% dario amodei approach 70%