US government orders halt on Anthropic's advanced AI models due to security concerns.
Anthropic's pre-IPO shares have taken a hit following a government directive to suspend access to its flagship AI models, Fable 5 and Mythos 5, citing national security risks associated with a potential jailbreak. The company's perpetual contract on Hyperliquid dropped approximately 3.7%, now trading around $1,627, down from over $1,800 shortly after Fable 5's launch.
On Friday, the U.S. government informed Anthropic at 5:21 p.m. ET to cease all access to its AI models for foreign nationals. In response, Anthropic complied but expressed strong discontent, arguing that the government's claims of a narrow vulnerability were overstated and not indicative of a comprehensive jailbreak. The company indicated that similar vulnerabilities are detectable by other models, including OpenAI's GPT-5.5.
Anthropic criticized the government's lack of formal evidence, stating that the risks presented were akin to routine software assessments that defenders undertake. The company warned that widespread implementation of such a directive could impede the deployment of new AI models across the industry.
