Mythos 5 access reopens after US security review ends

Mythos 5 regained limited access after US officials approved a partial restoration for select users, following talks over national security safeguards.

Anthropic PBC received the go-ahead from the US Commerce Department to resume sharing its advanced Mythos 5 artificial intelligence model with “certain trusted partners,” according to a letter signed by Commerce Secretary Howard Lutnick and viewed by Bloomberg News.

The decision marks a shift from restrictions imposed roughly two weeks earlier, when the US government moved to block foreign nationals from accessing Mythos 5 and a related model, Fable 5. That order led Anthropic to shut off worldwide availability to both systems while negotiations unfolded.

US lifts part of the ban after security talks

In the letter, Lutnick said Anthropic had coordinated with the US government on risks tied to what the department described as “Covered Models.” He wrote that the company’s work produced “significant progress,” enabling a limited release to a defined set of partners.

The earlier action was driven by concerns that security controls around the models could be bypassed, raising the prospect of misuse with national security implications. The government’s move focused on preventing foreign nationals from using the systems while officials assessed the strength of guardrails.

Commerce Department spokesman Benno Kass said the matter was handled on an accelerated timeline. “In just two weeks, we have worked diligently to ensure America remains the global leader in AI while safeguarding our security,” Kass said.

What changed, and what remains restricted

The Friday-dated letter signals a reopening of access for Mythos 5, but only under a narrower distribution approach. By limiting availability to “trusted partners,” the Commerce Department appears to be opting for a controlled rollout rather than broad public access.

The document does not indicate that restrictions have been eased for the Fable 5 model. As a result, the earlier limits affecting Fable 5 appear to remain in place, at least under the terms described in the correspondence.

Anthropic’s temporary global shutdown underscored how quickly regulatory decisions can affect access to frontier AI tools. It also highlights the leverage governments hold when access rules depend on compliance with security standards.

Why this matters for AI oversight and deployment

The episode illustrates an emerging pattern in US AI oversight: rapid intervention when officials believe advanced capabilities could be exploited, followed by negotiated technical and operational fixes. Rather than a permanent prohibition, regulators may use time-limited restrictions to pressure companies to strengthen controls.

It also reinforces the tension between global AI commercialization and national security screening. When restrictions are tied to user nationality or cross-border access, companies may respond with broader shutdowns to avoid compliance gaps while seeking clarity from regulators.

Next steps will likely center on how “trusted partners” are defined and audited, and whether similar conditions will be applied to other advanced systems. The status of Fable 5—and whether it can meet the threshold for reinstatement—will be a key indicator of how durable the current compromise is.