{"slug": "feds-legal-basis-for-ban-on-anthropics-most-powerful-models-looks-increasingly", "title": "Feds’ Legal Basis for Ban on Anthropic’s Most Powerful Models Looks Increasingly Shaky", "summary": "The Trump administration's ban on Anthropic's advanced AI models Fable 5 and Mythos 5 faces growing legal and expert scrutiny, with cybersecurity experts and legal scholars questioning the national security rationale. The ban, based on Export Administration Regulations, was imposed after Amazon CEO Andy Jassy reportedly briefed officials on the models' ability to identify cybersecurity flaws, but critics argue the move is misguided and could set a dangerous precedent for AI regulation.", "body_md": "The Trump administration’s ban on the use of Anthropic’s most advanced AI models—[Fable 5 and Mythos 5](https://gizmodo.com/anthropic-releases-a-safer-version-of-its-too-dangerous-mythos-ai-2000769492)—has been in effect for five days, with no clear end in sight. During that time, questions about the motives behind the directive have steadily grown.\n\nOstensibly, the government based the ban on national security concerns. Federal officials, including Treasury Secretary Scott Bessent, reportedly learned from Amazon CEO Andy Jassy last week about research conducted by the company showing that Anthropic’s Fable 5 was able to identify flaws in the company’s cybersecurity software. (Amazon has invested billions in Anthropic over the past three years and also provides some of the graphics processing units powering its family of Claude AI models.) Then, on Friday, Secretary of Commerce Howard Lutnick sent a letter addressed to Anthropic CEO Dario Amodei, which said the company was required to obtain an official government license before Fable 5 or Mythos 5 could be used by any “foreign persons” inside or outside the U.S., including Anthropic’s own employees.\n\nAnthropic was told it had less than ninety minutes to comply, sending the company into a [state of confusion](https://www.nytimes.com/2026/06/17/technology/anthropic-trump-administration-fable.html) and ultimately forcing it to cut off all users’ access to both models.\n\nLutnick’s letter didn’t mention anything about cybersecurity concerns or the meeting with Jassy. Instead, it cited the Department of Commerce’s Export Administration Regulations, or EAR, which gives the government authority to cut off the use of U.S.-made products or services if they pose “an unacceptable risk of use in, or diversion to, a ‘military-intelligence end use’ or a ‘military-intelligence end user,’” according to the letter to Amodei, which was first [published by Bloomberg](https://www.bloomberg.com/news/articles/2026-06-16/read-the-lutnick-letter-that-led-anthropic-to-disable-mythos). “Failure to comply will result in prompt criminal and civil penalties, as provided for by law,” the letter read.\n\n**Unanswered questions**\n\nFollowing the news about the meeting between Jassy and federal officials, a host of cybersecurity experts quickly [rallied to Anthropic’s side](https://gizmodo.com/cybersecurity-experts-are-baffled-by-trumps-ban-of-anthropics-new-ai-models-2000771976).\n\nAn open letter published Sunday and addressed to Lutnick and U.S. National Cyber Director Sean Cairncross argued that the government’s ban on Anthropic’s models was dangerously misguided. Not only does Fable 5—which, unlike Mythos 5, was made publicly available—come with uniquely robust safeguards, the letter said, but the ability to uncover cybersecurity vulnerabilities is not unique to Anthropic’s models. Many of the letter’s signatories “regularly use other foundation and open-source models for security audits and red-teaming every day,” it read. Anthropic also made this point in a blog post published in response to the government’s ban, arguing further that if this were to become the technical red line that AI developers were suddenly not allowed to cross, it would effectively outlaw the release of any new models made by any American tech company. The ban also makes it impossible for American cybersecurity professionals to use the models for defense purposes, the letter points out.\n\nSome legal experts have likewise found the government’s rationale for the ban tenuous. Lutnick’s brief letter to Anthropic “is so badly drafted it might not restrict API/chatbot access at all,” Alasdair Phillips-Robins, a former senior policy adviser for the Department of Commerce and current fellow in technology and international affairs at the Carnegie Endowment for International Peace, wrote in a [X post](https://x.com/alasdairpr/status/2067052688508899577) on Tuesday, following the letter’s publication. According to Phillips-Robins, users’ accessing Fable 5 or Mythos 5 through a digital interface doesn’t qualify as an “export,” and therefore should not fall under the purview of the EAR.\n\nEven if the federal government did have legitimate reason to deem Anthropic’s models as threats to national cybersecurity, the EAR and other laws governing the export of U.S.-made commodities “are not a roving license to ban unsafe products or punish companies the White House thinks are irresponsible,” Phillips-Robins continued. “If the administration fears dangerous AI models, it should work with Congress to write laws to govern them.”\n\nUnless and until Anthropic decides to challenge the decision in court, the Trump administration could use the justification to enforce a blanket licensing requirement across the industry. The administration has previously been reluctant to implement such a policy, favoring a more hands-off approach to AI. But in recent months, officials have been considering a system of federal control over the reach of AI models based on export laws, according to a Tuesday [report](https://www.theinformation.com/articles/anthropic-ban-stirs-concerns-openai-beyond-crackdown-foreign-ai-talent) from The Information.\n\n**The French connection**\n\nThe timing of the ban is also noteworthy. On Monday—three days after Lutnick’s letter was sent to Amodei—Trump arrived in the French resort town of Évian-les-Bains for the annual G7 Summit.\n\nThe ban of Fable 5 and Mythos 5 doesn’t just apply to geopolitical adversaries like Russia or Iran; it applies globally, meaning that even the U.S.’s closest allies are no longer able to use two of the most advanced AI models ever built. It could become a point of leverage for the U.S., a kind of digital blockade. Trump has already [refused ](https://www.telegraph.co.uk/business/2026/06/16/trump-snubs-starmer-plea-for-anthropic-ai-exemption/)British Prime Minister Keir Starmer’s request for an exemption to the ban.\n\nThe Trump Administration’s blacklisting of Fable and Mythos has also deepened anxieties among some European leaders about their countries’ dependence upon AI and other technologies owned by American companies. French President Emmanuel Macron met with Amodei and OpenAI CEO Sam Altman at Evian-les-Bains to discuss the future of European access to the companies’ most advanced models, Bloomberg [reported](https://www.bloomberg.com/news/articles/2026-06-17/macron-seeks-way-around-trump-s-ban-on-anthropic-s-ai-models?srnd=phx-politics).", "url": "https://wpnews.pro/news/feds-legal-basis-for-ban-on-anthropics-most-powerful-models-looks-increasingly", "canonical_source": "https://gizmodo.com/feds-legal-basis-for-ban-on-anthropics-most-powerful-models-looks-increasingly-shaky-2000773192", "published_at": "2026-06-17 20:05:21+00:00", "updated_at": "2026-06-17 20:24:48.588758+00:00", "lang": "en", "topics": ["ai-safety", "ai-policy", "ai-research", "ai-products", "ai-ethics"], "entities": ["Anthropic", "Fable 5", "Mythos 5", "Amazon", "Andy Jassy", "Scott Bessent", "Howard Lutnick", "Dario Amodei"], "alternates": {"html": "https://wpnews.pro/news/feds-legal-basis-for-ban-on-anthropics-most-powerful-models-looks-increasingly", "markdown": "https://wpnews.pro/news/feds-legal-basis-for-ban-on-anthropics-most-powerful-models-looks-increasingly.md", "text": "https://wpnews.pro/news/feds-legal-basis-for-ban-on-anthropics-most-powerful-models-looks-increasingly.txt", "jsonld": "https://wpnews.pro/news/feds-legal-basis-for-ban-on-anthropics-most-powerful-models-looks-increasingly.jsonld"}}