Yo team — drop EVERYTHING because Anthropic just pulled the plug on its Fable 5 and Mythos 5 models, shutting off access for every customer starting FRIDAY NIGHT. The timing is insane — these were launched only days earlier! This move came after a Commerce Department directive slapped export controls on them, restricting their use outside the US. Why? The administration's line is that there are jailbreak concerns around safeguards designed to block prompts about cybersecurity, chemistry, and biology, and they need time for the "national security apparatus" to be hardened against such threats — which should take a few weeks.
But Anthropic pushed back hard on how this was handled. They stated government officials gave them only verbal evidence of one narrow jailbreak involving Fable 5 reviewing a specific codebase for software flaws, and they've seen nothing beyond minor vulnerabilities in that scenario. Their argument is powerful: if every model facing a narrow potential exploit got pulled like this, it would essentially halt all new frontier deployments across the entire industry, since GPT-5.5 has similar capabilities on its own. Also worth noting — the butterfly imagery Anthropic used for Fable 5 is darkly ironic now that the model can't fly free!
This isn't just one bad call; there's a bigger picture here. Earlier this month President Trump signed an executive order pushing AI companies toward voluntary government security testing, which followed several months of internal administration disagreements about it — even an official signing ceremony was postponed because officials couldn't agree on the terms. Anthropic has since apologized for what they called "a misunderstanding" and promised more details within 24 hours of their Friday night post. The whole situation highlights how thin the line is between responsible security research and a complete halt on AI deployment in the US.
Source: https://arstechnica.com/ai/2026/06/anthropic-shuts-down-fable-mythos-models-following-trump-admin-directive/
But Anthropic pushed back hard on how this was handled. They stated government officials gave them only verbal evidence of one narrow jailbreak involving Fable 5 reviewing a specific codebase for software flaws, and they've seen nothing beyond minor vulnerabilities in that scenario. Their argument is powerful: if every model facing a narrow potential exploit got pulled like this, it would essentially halt all new frontier deployments across the entire industry, since GPT-5.5 has similar capabilities on its own. Also worth noting — the butterfly imagery Anthropic used for Fable 5 is darkly ironic now that the model can't fly free!
This isn't just one bad call; there's a bigger picture here. Earlier this month President Trump signed an executive order pushing AI companies toward voluntary government security testing, which followed several months of internal administration disagreements about it — even an official signing ceremony was postponed because officials couldn't agree on the terms. Anthropic has since apologized for what they called "a misunderstanding" and promised more details within 24 hours of their Friday night post. The whole situation highlights how thin the line is between responsible security research and a complete halt on AI deployment in the US.
Source: https://arstechnica.com/ai/2026/06/anthropic-shuts-down-fable-mythos-models-following-trump-admin-directive/