In a swift reversal, the AI startup Anthropic has pulled its most recent model, Claude Fable 5, from service after U.S. authorities issued a security order. The directive bars foreign‑national users, prompting the company to shut down both Fable 5 and its related Mythos 5 to ensure compliance.
Anthropic described the system as “too powerful” for public release and noted that no specific technical concerns were disclosed by the government. Instead, officials cited a potential jailbreak— a method that can override built‑in safeguards and reveal hidden vulnerabilities.
The model itself is a rival to OpenAI’s ChatGPT and Google’s Gemini, offering sophisticated natural‑language responses. Prior to launch, Anthropic restricted beta access to a handful of organizations, citing the risk that the program could exploit or hack underlying computer systems.
Critics have questioned the company’s claim that the model was “too powerful,” suggesting it may be overstated. Some analysts noted that similar vulnerabilities can surface on other public models without a jailbreak, undermining the security narrative.
This setback arrives amid a broader legal battle between Anthropic and the Pentagon, which has labeled the firm a supply‑chain risk—a designation rarely used against U.S. companies. A judge has so far ruled the Pentagon’s directive unenforced, allowing government agencies to continue using Anthropic’s tools during ongoing litigation.
The incident reflects growing tension around AI tools that push boundaries of autonomous reasoning. As the industry wrestles with safety standards, regulators and private firms must navigate new challenges that arise when cutting‑edge models intersect with national security concerns.
