The clash between Washington’s national‑security hawks and one of the world’s leading AI labs erupted into the open on Friday, when the U.S. government abruptly forced Anthropic to switch off its most powerful models worldwide.

Early June: Safety worries and commercial pressure

In early June, reporting highlighted Anthropic’s tight restrictions on its cutting‑edge systems, especially Mythos 5, which was being treated as a sensitive cybersecurity tool and shared only with about 50 vetted organizations under “Project Glasswing.”¹ At the same time, open‑source advocates and developers accused the company of quietly degrading its new general‑purpose model Fable 5 for advanced AI‑development tasks, routing such queries to a weaker system.²

Critics said this undercut claims that the limits were purely about national security and safety, arguing Anthropic was also protecting itself from competitors who could “distill” its outputs into rival models.² AI researcher Yann LeCun amplified those concerns, sharing a complaint that “as believers of open research, we are disappointed to see Anthropic silently degrading Fable 5 for AI development.”Yann LeCun (ylecun)³

June 12: Export controls and a Pentagon blacklist

On Friday, Commerce Secretary Howard Lutnick sent a letter placing Anthropic’s Mythos 5 and Fable 5 under strict export controls, barring access for foreign governments, firms, and individuals, including foreign nationals inside the U.S.¹ Anthropic was simultaneously added to a Pentagon blacklist labeling the models too risky for U.S. government use, even as Commerce required licenses for any domestic or international transfer of the systems.¹

Government shutdown order and Anthropic’s pushback

Later that day, at 5:21 p.m. ET, Anthropic received an export‑control directive requiring it to suspend access to Fable 5 and Mythos 5 for any foreign national. To comply, the company said it had to “abruptly disable Fable 5 and Mythos 5 for all our customers,” while other models remained online.⁴

The underlying trigger, officials told reporters, was an alleged jailbreak of Mythos that raised alarms about national‑security risks.¹ But Anthropic argued that what the government described was only a “potential narrow, non‑universal jailbreak” akin to getting the model to analyze a specific codebase for known, minor vulnerabilities—tasks it said other public models can also perform without bypasses.⁵⁴

The company publicly objected that this “narrow potential jailbreak” should not justify recalling models deployed to “hundreds of millions of people,” warning that such a precedent could stall progress across the AI industry.⁵

A broader fight over control of frontier AI

For the Trump administration, the move fits a broader strategy of treating cutting‑edge AI as a strategic asset, building on an earlier executive order to test advanced models before deployment.¹ For Anthropic and many developers, it raises a different fear: that national‑security concerns will increasingly be used to lock down AI capabilities, entrenching a small group of powerful providers while sidelining open research and open‑source competitors.²

"quotes": [ { "index": 1, "source": "web", "title": "Scoop: Trump admin blocks foreign access to Anthropic's most powerful AI", "text": "The Trump administration is blocking foreign governments, companies and individuals from accessing Anthropic's most advanced AI models — leading the company to cut off all customer access altogether.", "url": "https://www.axios.com/2026/06/12/anthropic-trump-mythos-fable-national-security", "authorPubkey": "a81a7ca9012c0794b5469398115d2c7a02b7833e21bb77fc19da7d4ae3f9bbd7", "profilePicture": null, "xHandle": null, "xId": null, "xUserName": null }, { "index": 2, "source": "web", "title": "This week's Anthropic-inspired AI freakout, explained", "text": "Anthropic said it will no longer secretly degrade Fable 5 responses when users ask for help on frontier AI model development ... But Anthropic is still restricting use of its most powerful public model for certain AI development work.", "url": "https://www.businessinsider.com/anthropic-freaked-out-ai-industry-mythos-fable-open-source-models-2026-6", "authorPubkey": "f37eb27bb6eb542573eb49dc7d3baa2e09ee022d5029e06440f561d8853923b5", "profilePicture": null, "xHandle": null, "xId": null, "xUserName": null }, { "index": 3, "source": "tweet", "title": "@ylecun on X", "text": "RT @askalphaxiv: As believers of open research, we are disappointed to see Anthropic silently degrading Fable 5 for AI development "Any to…", "url": "https://twitter.com/ylecun/status/2064827994636423215", "authorPubkey": null, "profilePicture": "https://pbs.twimg.com/profile_images/1483577865056702469/rWA-3_T7_normal.jpg", "xHandle": "Yann LeCun", "xId": "48008938", "xUserName": "ylecun" }, { "index": 4, "source": "web", "title": "Statement on the US government directive to suspend access to Fable 5 and Mythos 5", "text": "The US government ... has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national ... The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.", "url": "https://www.anthropic.com/news/fable-mythos-access", "authorPubkey": "378493c8987ffa5032103dd26dfa1d08c4a069d516551711ff3f013251d45aeb", "profilePicture": null, "xHandle": null, "xId": null, "xUserName": null }, { "index": 5, "source": "web", "title": "Anthropic's safety warnings may have just backfired — the government has pulled the plug on its most powerful AI", "text": "Anthropic says its understanding is that the underlying concern is a claimed jailbreak of Fable 5 ... a 'potential narrow, non-universal jailbreak' ... and argues this should not be cause for recalling a commercial model deployed to hundreds of millions of people.", "url": "https://techcrunch.com/2026/06/12/anthropics-safety-warnings-may-have-just-backfired-the-government-has-pulled-the-plug-on-its-most-powerful-ai/", "authorPubkey": "22a173d042acd0b75a45b80787fe1771dcbe171c8934cb343af7b219865595d5", "profilePicture": null, "xHandle": null, "xId": null, "xUserName": null } ]}