US Drops Export Controls on Anthropic’s Fable 5 and Mythos 5

The US Commerce Department has fully lifted export controls on Anthropic’s Fable 5 and Mythos 5 AI models, ending a freeze that began on June 12 over a reported jailbreak. Restoration of global access begins from Wednesday.

The reversal is conditional. Commerce Secretary Howard Lutnick told Anthropic in a letter seen by Reuters that the controls were withdrawn after Anthropic agreed to monitor its models for security risks, work with the government on release protocols, and notify Washington of malicious activity. The deal closes a standoff between the Trump administration and one of the country’s leading AI labs, and it puts Anthropic under the same Commerce Department review arrangement that OpenAI has already accepted for its own frontier release.

What Anthropic Committed to Washington

In the letter, Lutnick said Anthropic has appropriate safeguards in place and that “a license will no longer be required” to export or transfer the two models. The full text of the agreement, dated Tuesday, was not released publicly, but the obligations were detailed in news reports citing the letter.

Anthropic’s three commitments to Washington, per Lutnick’s letter:

• Monitor for security risks. Anthropic agreed to “proactively detect and address security risks associated with the models,” Lutnick wrote.
• Coordinate on protocols. The company committed to working “diligently with the US government on protocols and standards and releases for Mythos, Fable, and future models.”
• Report malicious activity. Anthropic must “inform the US government of any malicious activity” linked to either model.

The agreement covers “protocols and standards and releases” for “future models” beyond Fable 5 and Mythos 5, locking Anthropic into a standing consultation role with the department on every frontier release, according to the underlying Commerce letter. The original June 12 directive remains formally in place but unused; Lutnick wrote in the partial-release letter dated June 26 that Commerce may still “reevaluate and adjust the scope of license requirements on” Anthropic’s models “should circumstances change.”

The June 12 Directive, Line by Line

The freeze began at 5:21pm ET on Friday, June 12, when the Commerce Department’s Bureau of Industry and Security delivered an export-control directive to Anthropic. The order was sweeping: it suspended access to Fable 5 and Mythos 5 “by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.”

Anthropic complied within hours, disabling both models for all customers worldwide, per Anthropic’s statement on the directive. The order did not specify the security concern in writing, but the company’s own statement said the government believed it had “become aware of a method of bypassing, or ‘jailbreaking’ Fable 5.” A senior White House official told Politico the Commerce letter that triggered the shutdown was a “last resort after begging them for hours to work with us.”

Anthropic’s position has been that “perfect jailbreak resistance is not currently possible for any model provider,” and that “other publicly-available models are able to discover” the same kinds of vulnerabilities. The company said it had reviewed a demonstration of the technique used in the directive and concluded the underlying capability was “widely available from other models (including OpenAI’s GPT-5.5).”

The key dates of the standoff:

1. April 2026: Anthropic shares a limited preview of Mythos with cybersecurity partners.
2. June 9, 2026: Anthropic launches Fable 5 and Mythos 5 publicly.
3. June 12, 2026: Commerce Department directs Anthropic to disable both models for all foreign national access; Anthropic complies the same evening.
4. June 26, 2026: Commerce partially lifts the freeze, clearing Mythos 5 for a group of about 100 trusted US organizations.
5. June 30, 2026: Commerce fully lifts the controls on both Fable 5 and Mythos 5; Anthropic says access will be restored from July 1.

From 100 Trusted Partners to Full Restoration

The full lifting followed a two-stage rollback. On Friday, June 26, Lutnick’s office sent a letter to Anthropic co-founder Tom Brown clearing Mythos 5 for release to a group of trusted US companies and federal agencies, in what was the partial US approval Anthropic won for Mythos 5. The partial release did not cover Fable 5, and Lutnick told Anthropic that Commerce “may amend the list of entities at any time.” Tom Brown, not CEO Dario Amodei, took the lead in negotiations with the Trump administration, CNBC reported.

Anthropic’s Project Glasswing cybersecurity initiative already had US tech firms signed up as of early June, including Apple, Alphabet’s Google, Cisco Systems, Nvidia, and Microsoft, Fortune reported. Glasswing members were cleared to use Mythos for defensive cybersecurity tasks under the original release plan.

The numbers behind the deal:

• ~100 companies and federal agencies initially cleared for Mythos 5 access.
• ~200 firms were already in Anthropic’s Project Glasswing cybersecurity program.
• ~80 cybersecurity executives signed an open letter to Lutnick asking for the restrictions to be lifted.

The Finding That Triggered the Ban

Anthropic believes the directive was based on a single jailbreak demonstration that Amazon CEO Andy Jassy shared with the government, according to The Wall Street Journal. Anthropic’s own assessment was that the technique was narrow, used to identify “a small number of previously known, minor vulnerabilities.” Amazon is an Anthropic investor.

The cybersecurity industry pushed back hard. A group of nearly 80 cybersecurity executives and experts sent an open letter to Lutnick and National Cyber Director Sean Cairncross asking the White House to lift the restrictions and “commit to an open, scientific and transparent process of handling AI risk assessments in the future.” The letter warned that pulling “the best capabilities away from defenders without a good reason when our adversaries are rapidly advancing is dangerous,” and that Chinese open-weight models “are only months behind the best American models.”

Former White House AI czar David Sacks supported the Trump administration’s move. “The Admin values Anthropic’s technical capabilities and feels that this issue, while serious, should be easily resolved,” he said. “The ball is in Anthropic’s court.” Policy analysts documented the European reaction in wider analysis of the directive’s global implications, with former French interior minister Bruno Retailleau calling it “a wake-up call.”

How the Same Arrangement Reached OpenAI

The deal’s structure matters beyond Anthropic. On the same day the partial release landed, OpenAI released its GPT-5.6 family of models under similar conditions. The rollout was limited to “a small group of trusted partners whose participation has been shared with the government,” the company said.

OpenAI framed the move as a temporary step. “We don’t believe this kind of government access process should become the long-term default,” the company wrote. “It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.” OpenAI previewed the models with the government before launch.

If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.

That quote is from Anthropic’s statement on the day it disabled both models. The Lutnick letter requires the same coordination for future releases.

OpenAI CEO Sam Altman publicly endorsed the safety testing behind the arrangement. He drew a line at the idea of the government picking customers. The distinction matters: OpenAI’s release was voluntary, while Anthropic’s was the result of an export-control order.

The Two Fights Anthropic Still Has

The export controls are gone, but the broader fight between Anthropic and the Trump administration is not. In March, the Defense Department declared Anthropic a supply-chain risk, a label historically reserved for foreign adversaries, after contract talks over military use of its models broke down. Anthropic has sued the administration over the blacklisting, and litigation is ongoing. The designation blocks Anthropic from working with defense contractors on military AI.

Anthropic also filed confidentially for an initial public offering shortly before the June 12 directive, with its last private valuation topping $900 billion, per Fortune’s reporting. Anthropic has not commented on the timing relative to the Commerce directive. The Commerce Department’s monitoring arrangement does not affect the Defense Department’s separate designation.

For now, the deal is on. “We’ll begin restoring access tomorrow,” Anthropic said Tuesday night. From Wednesday, both Fable 5 and Mythos 5 are back for every user, foreign nationals included, subject to the obligations Anthropic accepted in the Lutnick letter.

Frequently Asked Questions

What are Fable 5 and Mythos 5?

Fable 5 is Anthropic’s first public-facing model from its Mythos cybersecurity line, with built-in safeguards that block responses on certain cybersecurity and biology queries. Mythos 5 is a less-restricted version of the same model distributed only to a vetted group called Project Glasswing.

Why did the US Commerce Department restrict the models?

The agency issued the June 12 directive after the government concluded a method existed to bypass Fable 5’s safeguards. Anthropic has argued in its own statement that the technique was narrow and that other publicly available models can find similar vulnerabilities. The original directive did not specify the security concern in writing.

What did Anthropic have to concede to get the models restored?

Three obligations, per Lutnick’s letter: proactively detect and address security risks, work with the US government on protocols for future releases, and notify the government of any malicious activity linked to the models. The original June 12 directive remains formally in effect.

Who now has access to Fable 5 and Mythos 5?

Both models are being restored for all users, foreign nationals included. The earlier partial release had cleared Mythos 5 for about 100 trusted US companies and federal agencies, with Apple, Google, Cisco, Nvidia, and Microsoft among the existing Project Glasswing partners.

Can the export controls come back?

Yes, the framework is dormant rather than repealed. The June 12 directive is still formally in effect, and Lutnick’s letter states that Commerce can revisit the scope of license requirements at any time. The cleared-entities list can also be amended without notice.