Anthropic reopened access to its advanced AI models Fable and Mythos on Wednesday, following the Trump administration's lifting of export control restrictions.

According to anofficial announcement, the return of these two models—Fable 5, available to the general public, and the more powerful Mythos 5, restricted to a trusted partner alliance, consistent with the pre-ban arrangement—marks significant progress in negotiations between the Trump administration and AI companies over the responsible deployment of frontier AI models. In recent months, the U.S. government and the tech industry have been at odds on this issue.

In astatementreleased on Tuesday, Anthropic announced a resolution to the standoff, insisting that its models have always been safe and arguing that the government overstated the risks. This stance suggests that tense conversations lie ahead over how to balance providing advanced capabilities to defenders while preventing U.S. adversaries from obtaining them.

The U.S. Department of Commerce imposed the export control ban after Amazon warned the government that Fable's safety guardrails could be bypassed. On Tuesday, Anthropic reiterated its argument—also supported by acoalition of leading cybersecurity experts—that less capable AI models face the same issues, but the company also said it had "acted quickly to address the bypass issues mentioned in the report."

Anthropic said that over the past two weeks, it worked closely with "the government and other partners, including Amazon," and "trained an improved safety classifier to identify and block the behaviors described in the report."

Anthropic added that researchers at the U.S. National Institute of Standards and Technology's (NIST) AI Standards and Innovation Center "have tested our previous and new safety guardrails and found them all to be robust."

Meanwhile, the company warned that these changes could have some negative effects on cybersecurity researchers seeking assistance with defensive work.

"The new classifier... will flag more benign requests in routine coding and debugging tasks, at the cost of increased false positives," Anthropic said. "As with all safety guardrails, we will continue to refine it to better distinguish genuine abuse from legitimate requests and reduce false positives."

Calls for a more formal review process

Although the controversy surrounding Fable and Mythos may have ended, the AI industry remainsconcerned。

about the arbitrary manner in which the Trump administration reviews the availability of frontier models. Trump recentlysigned an executive orderestablishing a process for frontier AI companies to provide the government with early access to certain particularly powerful models. On Wednesday, Anthropic said it would provide such early access for "models that substantially advance the capability frontier in areas relevant to national security." The company also said it would share intelligence on how hackers might abuse its tools and participate in the vulnerability information-sharing center established under Trump's directive.

Anthropic emphasized its commitment to working closely with the government to address potential AI safety risks. The company said it is "significantly expanding" collaboration with federal agencies, including dedicating specialized personnel and computing resources. It also pledged to "work with government and industry peers to jointly develop voluntary safety and evaluation standards for frontier model providers."

Anthropic also noted that there is currently "no recognized standard" for classifying the severity of jailbreak attacks—a critical prerequisite for any formal model review process.

"A common standard for evaluating AI jailbreak attacks would help us and other companies safely release new models while also allowing users to fully leverage their advanced capabilities," Anthropic said.

To that end, the company announced it is working with Amazon, Google, Microsoft, and other members of Project Glasswing—through which Anthropic grants Mythos access to vetted organizations—to develop a "consensus framework" for classifying and responding to jailbreak attacks. The company said the framework envisions rating each potential jailbreak against four criteria, including the difficulty of discovering the bypass method and the additional model capabilities unlocked by the bypass.