On Friday night, a federal order compelled Anthropic to block its flagship generative‑AI models, Fable 5 and Mythos 5, from any foreign user, regardless of location. The mandate, issued on national‑security grounds, extended to Anthropic employees, effectively halting all external access to the two systems.
Anthropic’s response was swift. Within hours, the company disabled the models for every paying customer, citing compliance with the government’s directive. In a concise press release, the firm noted that the order arrived without detailed justification, leaving it to interpret vague concerns about a possible "jailbreak" vulnerability.
According to Anthropic, the government’s claim rested on a verbal briefing rather than a formal written report. The company said the alleged exploit was minor, already observable in other large‑language models, including the rumored GPT 5.5. Anthropic emphasized that it had never received a concrete disclosure of a non‑universal jailbreak that resulted in harmful output. The few findings it had seen were either harmless or offered only negligible advantage to Mythos‑specific tasks.
To address the issue, Anthropic outlined a series of mitigation steps. It has been coordinating with both U.S. and U.K. authorities to monitor attempts to misuse the models. The firm also revised its data‑retention policy, aiming to log interactions that could signal malicious intent. These measures, Anthropic said, are designed to help trace and prevent future exploitation.
Industry observers note that the order underscores growing tension between rapid AI development and governmental oversight. While Anthropic’s models are among the most advanced in the market, the lack of public detail about the security threat fuels speculation about the scope of the vulnerability and the potential impact on other AI providers.
For customers, the shutdown means an abrupt loss of access to tools that many had integrated into products ranging from content creation to code assistance. Anthropic has not indicated a timeline for reinstating the models, pending clarification from officials.
The episode adds to a mounting list of regulatory actions aimed at curbing AI risks, from export controls to mandatory safety audits. As policymakers grapple with the balance between innovation and security, companies like Anthropic find themselves navigating an increasingly complex legal landscape.
Dieser Artikel wurde mit Unterstützung von KI verfasst.
News Factory APP - agentische News für besseres SEO & AEO.