The Trump administration has forced AI company Anthropic to take its new Fable 5 and Mythos 5 models offline using an unprecedented export control order. The move was triggered by a potential "jailbreak" vulnerability reported by investor Amazon, but Anthropic claims the flaw is minor. Critics allege the action is politically motivated retaliation, citing a pre-existing conflict with the Pentagon, and warn it creates an "ad hoc" regulatory regime that risks US AI leadership.
Fable export control ban
- ▪The order was a federal export control directive from the U.S. Commerce Department, banning all foreign nationals from using the models
- ▪Administration officials reportedly gave Anthropic a 90-minute deadline to voluntarily take down the Fable 5 model
- ▪To comply with the directive, Anthropic shut down access to the models for all users, including domestic customers and its own non-US citizen employees
- ▪The Trump administration ordered Anthropic to take its Fable 5 and Mythos 5 AI models offline, citing national security concerns
Amazon jailbreak research
- ▪Anthropic, which has received a $13 billion investment from Amazon, argued the potential jailbreak was minor and did not warrant pulling the model
- ▪The government's action was triggered after Amazon CEO Andy Jassy shared research with the White House detailing a potential "jailbreak" of Fable 5's safety features
- ▪The National Security Agency (NSA) concluded that there are ways to disable the safety guardrails on the Fable 5 model
- ▪Cybersecurity expert Katie Moussouris stated the jailbreak was not a novel risk and that other models like OpenAI's GPT-5.5 have similar capabilities
Political motivation allegations
- ▪Former Trump official Taylor Budowich stated the issue was not political, but that Anthropic "abandoned its safety guardrails" and its partner Amazon "felt compelled to blow the whistle."
- ▪The Trump administration was reportedly considering sanctions on Anthropic weeks before the order, due to a dispute over the company sharing technology with a suspected China-linked firm
- ▪Critics suggest the order was politically motivated, citing a "fractious relationship" and "ideological differences" between the administration and Anthropic
AI safety regulatory process
- ▪Dozens of cybersecurity experts signed an open letter urging the government to lift the controls, arguing the ban removes key tools from cyber defenders and risks US AI leadership
- ▪Experts noted the use of export controls is unprecedented for a remotely accessed AI service where no software or model weights are transferred to the user
- ▪The Trump administration believes Anthropic should be more proactive in continually testing its models for jailbreaks and flagging them to the government
- ▪Policy advocates criticized the move as an "ad hoc" approach to AI regulation, establishing a de facto licensing regime without clear standards
Anthropic-Pentagon conflict background
- ▪The Pentagon's designation followed a dispute over Anthropic's refusal to allow its models to be used for mass surveillance or fully autonomous weapons
- ▪Anthropic is currently challenging the Pentagon's "supply chain risk" designation in court
- ▪Earlier in 2026, the Pentagon, under Defense Secretary Pete Hegseth, designated Anthropic a "supply chain risk."
Story comments
Loading comments…