Anthropic CEO Dario Amodei is set to meet Commerce Secretary Howard Lutnick on Monday in Washington over the future of Fable 5 and Mythos 5, the company’s most advanced AI models, two people familiar with the matter told the BBC. The meeting is the first face-to-face since the Commerce Department issued an export control directive on June 12, 2026 that forced Anthropic to suspend all access to the models. Anthropic is pressing to bring both models back online.
A surprise third party has joined the fight. Dozens of cybersecurity executives, including former Facebook and Yahoo chief security officer Alex Stamos, have signed an open letter to Lutnick arguing the export control strips US defenders of their best tool at the worst possible moment. The letter, reported by the BBC and others, frames the directive as a self-inflicted wound for American AI leadership.
The Monday Meeting on Fable 5
The Monday session is set for Washington DC between Anthropic executives and the Department of Commerce, which Lutnick leads, per two people familiar with the planning. The Commerce talks are expected to include more documentation of the alleged vulnerability that triggered the order. It is unclear whether Anthropic will be able to restore access to either model after the meeting.
No comment from the White House. Commerce and Anthropic representatives did not respond to a request for comment. The BBC first reported the meeting.
Delivered to Anthropic on June 12, the Commerce order requires the company to bar any foreign national from using Fable 5 or Mythos 5, including foreign national Anthropic employees, per The official response to the June 12 export control. To comply, Anthropic disabled both models for all customers worldwide. The directive marked an escalation in Washington’s effort to treat cutting-edge AI as a national security asset, per Axios.
How the June 12 Directive Hit Anthropic
The Commerce Department letter, signed by Lutnick and delivered at 5:21 p.m. ET on June 12, did not include specific details of the national security concern, Anthropic said. The “net effect” of the order, the company wrote, was to “abruptly disable” the two models “to ensure compliance.” To comply, Anthropic pulled both models within hours of receiving the letter.
Fable 5 had launched publicly only three days earlier, on June 9, as Anthropic’s first Mythos-class model made available for general use, per The June 9 launch announcement for Fable 5. Mythos 5, the more capable sibling, was reserved for a select group of cyber defenders and infrastructure providers under a program called Project Glasswing. Both models had been reviewed by the US government and the UK’s AI Security Institute before the public release, per Politico.
The trigger for the directive was Amazon CEO Andy Jassy raising concerns to the White House on Thursday June 11 about the ability to bypass Fable 5’s guardrails, per How a CEO warning led to the Fable 5 export control. The National Security Agency reviewed Amazon’s findings and concluded they amounted to “proof,” a White House official told Politico. Anthropic said on Friday it had received only “verbal evidence” of the purported jailbreak.
- April 2026: Anthropic launches Project Glasswing and releases Claude Mythos Preview to a limited group of cyber defenders.
- May 2026: Anthropic and senior White House officials hold “productive” sessions leading to an executive order on voluntary AI pre-deployment testing.
- June 9, 2026: Anthropic launches Fable 5 publicly and Mythos 5 to selected partners.
- June 11, 2026: Amazon CEO Andy Jassy raises Fable 5 guardrail-bypass concerns to the White House.
- June 12, 2026, 5:21 p.m. ET: Commerce Department delivers the export control directive to Anthropic.
- June 12, 2026, late: Anthropic suspends both models for all customers.
- June 13, 2026: Cybersecurity executives publish an open letter to Lutnick and Cairncross.
- June 15, 2026: Anthropic’s team is set to meet Commerce Department officials in person in Washington.
Anthropic Calls the Jailbreak “Narrow”
Anthropic has accepted the directive. The company is contesting the premise. “No testers have yet been able to find a universal jailbreak” that would broadly bypass the model’s safeguards, the company wrote on June 12.
The vulnerabilities it has reviewed, Anthropic added, are “minor findings that provide no Mythos-specific uplift.” Anthropic said the level of capability in the government report is “widely available from other models (including OpenAI’s GPT-5.5), and is used every day by the defenders who keep systems safe.”
The signatories of the open letter to Lutnick and National Cyber Director Sean Cairncross made the same point: the technique “can be replicated on GPT-5.5, Opus, Sonnet and even Chinese models like Kimi 2.7,” per the New York Post. Anthropic has called perfect jailbreak resistance “not currently possible for any model provider” and has built a defense-in-depth strategy around the fact that some jailbreaks will always get through. The company said the launch was preceded by thousands of hours of red-teaming with the US government, the UK AISI and “multiple private third-party organizations.”
Fable 5 launched at $10 per million input tokens and $50 per million output tokens, less than half the price of Claude Mythos Preview, per The June 9 launch of Fable 5 as a Mythos-class model. Its safeguards trigger on average in less than 5% of sessions, Anthropic said, and were tuned to be “conservatively” cautious. If the standard for pulling a model is “a narrow potential jailbreak,” Anthropic wrote, “we believe it would essentially halt all new model deployments for all frontier model providers.”
We believe this is a misunderstanding and are working to restore access as soon as possible.
Anthropic said in its June 12 statement on the export control directive. The statement added that the company had complied with the order within hours of receiving it.
- Fable 5 launch date: June 9, 2026
- Input price: $10 per million tokens
- Output price: $50 per million tokens
- Safeguard trigger rate: less than 5% of sessions
- Open letter signatories: dozens (per the BBC), nearly 80 (per the IAPP)
The Cyber Defenders’ Open Letter
The open letter, addressed to Lutnick and Cairncross, was a rare case of the cybersecurity community publicly siding with an AI lab against a national security directive. The signatories, dozens per the BBC and nearly 80 per the IAPP, include former Facebook and Yahoo chief security officer Alex Stamos (now chief product officer at Corridor), Katie Moussouris (CEO of Luta Security), Rachel Tobac of SocialProof Security, Chris Wysopal of Veracode and Joe Levy (CEO of Sophos). Other signatories include security staff from Nvidia, Zoom, Mercedes-Benz and former US government and Google security teams, per the BBC.
The letter argues that pulling the most capable AI from US hands “without a good reason” strips defenders of their best tool at the worst moment. Chinese open-weight models, the signatories wrote, “are only months behind the best American models.” The export control, in their framing, is an own goal: it leaves adversaries free to keep building while US cyber defenders fall back to weaker models. “To pull the best capabilities away from defenders without a good reason when our adversaries are rapidly advancing is dangerous,” the letter says.
The letter argues that the directive has weakened US cyber defenses and created market uncertainty without a real risk to justify it, per the BBC. The letter also asks the government to “commit to an open, scientific and transparent process of handling AI risk assessments in the future.”
This action has taken the best models away from defenders, created market uncertainty, and risked America’s AI leadership without any real risk to justify it.
The letter was addressed to Commerce Secretary Howard Lutnick and White House National Cyber Director Sean Cairncross, per the BBC. The signatories include security staff from Adobe and Zoom, plus former US government and Google security staff, per the New York Post. Adam Thierer, a senior policy analyst at the R Street Institute, called the Trump administration’s action a “significant escalation in the politicization of AI and centralization of control over advanced computation in this country” in a post on X, per the New York Post.
A Wider Standoff With Washington
The Monday meeting is the latest in a year-long deterioration of Anthropic’s relationship with the federal government. On March 3, Defense Secretary Pete Hegseth and the Pentagon designated Anthropic a “supply chain risk” over the company’s refusal to allow its models to be used for mass domestic surveillance or autonomous weapons. Anthropic sued the Defense Department in March over the designation.
Tensions eased in mid-May, when Anthropic and senior White House officials met in what both sides called “productive” sessions, per Politico. Those meetings fed a recent executive order asking frontier AI companies to voluntarily submit their most capable models to the government before deploying them. Fable 5 was the first major test of that voluntary regime.
The Commerce directive ran outside that voluntary framework. It was an export control, not a request for pre-deployment review, and Anthropic has not been given specifics of the jailbreak finding. Politico reported the administration initially tried to get Anthropic to pause the release of Fable 5 voluntarily; Anthropic rebuffed the request. “Export controls were a last resort after begging them for hours to work with us,” a senior White House official told Politico. Anthropic disputed that account: “The White House gave 90 minutes to take the models down, with no details on the actual threat,” a person close to the company said after publication.
What a Resolution Would Settle
If Anthropic convinces Lutnick’s team that the jailbreak is narrow and reproducible on other models, the export control could be lifted and Fable 5 returned to general release. The company has hinted it will publish further technical detail in the coming days. If the government holds the line, Anthropic’s most powerful publicly available model stays offline while the company races toward an initial public offering ahead of rival OpenAI. The dispute is also a stress test of the executive order’s voluntary testing regime.
Amodei is set to travel to France this week for the G7 conference, per the New York Post. The Monday meeting in Washington will happen before that trip. The Monday outcome will determine whether Fable 5 returns to general release.
Frequently Asked Questions
What are Fable 5 and Mythos 5?
Fable 5 is Anthropic’s first publicly available Mythos-class model, launched on June 9, 2026 with safety guardrails that route sensitive queries to a less-capable sibling model, Claude Opus 4.8. Mythos 5 is the same underlying model with safeguards lifted in some areas, reserved for a select group of cyber defenders and infrastructure providers under Project Glasswing. Both launched at $10 per million input tokens and $50 per million output tokens.
What did the Commerce Department directive do?
The Commerce Department, citing national security authorities, issued an export control directive on June 12, 2026 ordering Anthropic to suspend all access to Fable 5 and Mythos 5 by any foreign national, inside or outside the US, including foreign national Anthropic employees. To comply, Anthropic disabled both models for all customers worldwide. Per Axios, the directive marked an escalation in Washington’s effort to treat cutting-edge AI systems as national security assets.
Why does Anthropic say the jailbreak finding is wrong?
Anthropic says the government has only given it verbal evidence of a single, narrow, non-universal jailbreak that consists of asking the model to read a specific codebase and fix software flaws. The same technique, the company and the open letter signatories say, can be replicated on OpenAI’s GPT-5.5, Anthropic’s own Opus and Sonnet, and Chinese models including Kimi 2.7. Anthropic argues that no frontier model can be made fully jailbreak-proof and that the directive, applied as a standard, would halt all new model deployments.
Who signed the open letter to Lutnick and Cairncross?
Signers include former Facebook and Yahoo chief security officer Alex Stamos (now chief product officer at Corridor), Katie Moussouris (CEO of Luta Security), Rachel Tobac of SocialProof Security, Chris Wysopal of Veracode and Joe Levy (CEO of Sophos). The letter, addressed to Commerce Secretary Howard Lutnick and White House National Cyber Director Sean Cairncross, was signed by dozens of signatories per the BBC and nearly 80 per the IAPP. Other signatories include security staff from Adobe, Zoom, Nvidia, Mercedes-Benz and former US government and Google security teams.
What is the wider context of Anthropic’s dispute with Washington?
The Fable 5 dispute is the latest flashpoint in a year-long deterioration of Anthropic’s relationship with the federal government. In March 2026, the Defense Department under Secretary Pete Hegseth designated Anthropic a supply chain risk over the company’s refusal to allow its models to be used for mass surveillance or autonomous weapons. Anthropic sued the Defense Department in March, a case still working through the courts. A series of meetings in May between Anthropic and senior White House officials was described as productive on both sides and led to a recent executive order on voluntary AI pre-deployment testing. The Fable 5 directive sits outside that voluntary framework.
