150 Cybersecurity Leaders Urge Trump to Revoke Anthropic's Fable and Mythos Ban

More than 150 cybersecurity executives signed an open letter on Tuesday (16) addressed to U.S. Secretary of Commerce Howard Lutnick, requesting the immediate reversal of the ban that blocked access to Fable 5 and Mythos 5 from Anthropic. The document, hosted on freefable.org, includes signatories such as Alex Stamos, former CSO of Facebook; Katie Moussouris, founder of Luta Security; Casey Ellis, founder of Bugcrowd; Jon Callas, former Apple; Paul Vixie; Dino Dai Zovi, former Block; and Rachel Tobac, CEO of SocialProof Security.

The order from the Department of Commerce, released on Friday (12), prohibits Anthropic from providing the two models to any foreign entity, whether on American soil or not, without a specific export license. The measure forces Anthropic to disable Fable 5 and Mythos 5 for all users, including its own non-American employees. Bloomberg published the full text of Lutnick's letter to CEO Dario Amodei on Tuesday (16), stating that civil and criminal sanctions would be imposed for non-compliance. This was the first invocation of this obscure authority under the Bureau of Industry and Security for general-use AI.

The Technical Argument Against the Ban

The central thesis of the signatories is that removing Fable and Mythos from the market does not close any doors to an adversary and removes tools that defenders effectively use. "We banned the models from exactly the wrong side of the equation," said Katie Moussouris to Fortune. For Stamos, who led Yahoo's response to the attack attributed to Russian intelligence in 2014, the jailbreak that prompted the order does not constitute a unique threat: "Any adversary capable of operating at that layer has already been using Chinese open source for months," he wrote in a text accompanying the letter.

The directive stems from a private paper by Amazon researchers, according to TechCrunch and Axios. The authors demonstrated that by requesting "fix this code" instead of "review for security issues," Fable produced patches that opened exploitable vectors. Amazon CEO Andy Jassy communicated directly with the White House, bypassing the internal structure of Anthropic, according to BigGo Finance's investigation. Amazon has invested $13 billion into Anthropic since 2023 and secured an additional $20 billion investment.

The Message for Corporate Clients

The order creates three immediate operational problems for corporate users. The first is continuity. Banks such as BNP Paribas, Lloyds, MUFG, and Santander Spain, which were already using Mythos 5 through Amazon Bedrock for tasks such as reconciling structured operations, have been removed from the product without a migration window. The second is jurisdiction. The order applies to any foreign entity, anywhere in the world, directly affecting the offshore operations of consultancies that resold the model, such as Accenture and Capgemini.

The third is precedent. For the first time, the Department of Commerce imposed export control on an AI model outside the hardware layer. Until now, the export controls regime covered GPUs. It now covers weight. For Mistral in France and DeepSeek in China, the effect is the opposite of what was intended: both are seen as safer alternatives for global CIOs, with Mistral reportedly experiencing a traffic spike in its API between Saturday and Sunday, according to public data from CloudFlare Radar.

The European Reaction and the G7 Angle

European diplomats raised the issue with Lutnick in Évian-les-Bains during the G7 summit, according to the New York Post. The request was for "trusted partner" status for the United Kingdom, Germany, France, and Japan. A Trump administration official responded that exempting any country, even an ally, would be "completely illogical." This stance contrasts with the trilateral AI agreement signed by Macron, Merz, and Starmer in May, which anticipated cooperation in model security with Anthropic itself.

The relevant fact for the coming week is the progress of the letter. The signatories are requesting a meeting with the National Cyber Director and cite technical exchanges that are already taking place between the Anthropic team and the Department of Commerce. If the White House retreats, it opens a window for a "general license" regime used in cryptography in the 1990s when the same Bureau ended up releasing PGP after similar pressure from researchers. If it holds, the American AI sector will begin to operate under model-specific export restrictions, and the European Commission gains an explicit argument to accelerate funding for sovereign models within the AI Continental Plan.