US reverses export restrictions on Anthropic’s Fable 5, Mythos 5 AI models

The US government has reversed export restrictions on Anthropic’s frontier AI models Fable 5 and Mythos 5, allowing the company to resume global access after nearly three weeks of disruption triggered by concerns over the models’ cybersecurity capabilities.

“As of today, June 30, the export controls on Fable 5 and Mythos 5 have been lifted,” Anthropic said in a blog post. The company said Fable 5 will begin rolling out globally on July 1 across Claude Platform, Claude.ai, Claude Code and Claude Cowork, while access on Amazon Web Services, Google Cloud, and Microsoft Foundry will be restored “as quickly as possible.”

Commerce Secretary Howard Lutnick said the administration had worked with Anthropic before reversing the restrictions.

“Over the past two weeks, we have worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the US Government and strengthen America’s leadership in AI,” Lutnick wrote in a post on X.

Anthropic had launched Fable 5 and Mythos 5 on June 9, and the US authorities restricted their export on June 12.

The Anthropic decision comes as other frontier AI developers are also operating under closer government scrutiny.

In announcing GPT-5.6 Sol, Terra, and Luna last week, OpenAI said it had previewed both its rollout plans and the model’s capabilities to the US government before launch and, “at their request,” was initially making the model available only to “a small group of trusted partners” whose participation had been shared with the government.

OpenAI said it would continue coordinating with government partners before expanding availability, but added that it did not believe “this kind of government access process should become the long-term default” because it keeps advanced AI tools from “users, developers, enterprises, cyber defenders, and global partners who need them.”

Analysts say the Anthropic decision reflects a broader shift in how frontier AI models are governed.

“The reversal is not the story; the instrument beneath it is,” said Sanchit Vir Gogia, chief analyst at Greyhound Research. Washington, he said, has applied the long-standing “deemed export” doctrine to frontier AI models, signalling “negotiated oversight, conditional and monitored, rather than blanket prohibition.”

Immediate order created global disruption

Anthropic said the June 12 export controls required it to restrict access to foreign nationals.

“Because the order took effect immediately and we had no reliable way to verify nationality in real-time, we suspended access to both models for all users,” it added. The company partially restored Mythos 5 last week for a limited group of US organizations.

Anthropic said the restrictions followed a report from Amazon researchers describing a technique that bypassed one of Fable 5’s cybersecurity safeguards. After reviewing the findings with Amazon and the US government, the company concluded the technique “did not expose any unique Mythos-level cyber capabilities” and instead represented “a borderline case for Fable 5’s safeguards.”

It subsequently retrained its safety classifier, saying the reported technique is now blocked in more than 99% of cases, while acknowledging that the stronger protections would increase false positives for some legitimate coding requests. Researchers at the US Department of Commerce’s Center for AI Standards and Innovation also evaluated the updated safeguards, Anthropic said.

Enterprises face a new continuity risk

Experts say the episode demonstrates that frontier AI’s availability can now be shaped by policy decisions as much as technical capability.

“Frontier access has become conditional infrastructure,” Gogia said. “The models went dark globally because nationality could not be verified in real time, so a control aimed at foreign nationals became an outage for everyone.”

He said enterprises should also not assume that deploying models across multiple cloud providers insulates them from regulatory actions affecting the underlying model provider.

According to Gogia, organizations should begin evaluating frontier AI platforms for regulatory interruption, cross-border identity restrictions, channel restoration delays, and trusted-partner eligibility alongside traditional security, commercial, and technical considerations.

Anthropic proposes a common jailbreak framework

Anthropic used the announcement to call for an industry-wide framework to assess AI jailbreaks, saying developers and governments currently lack a common standard for evaluating newly discovered techniques.

“There’s currently no consensus in the AI industry on how to describe, in objective terms, the severity of an AI jailbreak,” the company said. Anthropic said it is working with Amazon, Microsoft, Google, and other Project Glasswing partners on a framework for evaluating jailbreaks, while also expanding collaboration with the US government through pre-release testing of future frontier models, information sharing, and joint AI security research.

“Government involvement in AI releases requires a durable, transparent process that gives cyber defenders and others the certainty they need about access to powerful models,” Anthropic said. “These rules should be codified in strong regulation and applied equally across frontier model developers.”

Gogia said the broader lesson extends beyond Anthropic’s restored access. “Restored access is not restored certainty,” Gogia said. “Build for the detour, because the road now runs through policy.”