The global artificial intelligence (AI) industry faced fresh volatility as secondary market trading for Anthropic‘s shares took a hit. This followed a swift directive from US officials compelling the firm to restrict its latest high-performance models due to perceived security risks. The incident highlights escalating oversight of so-called frontier technologies amid recent national security considerations.
Late on Friday, US authorities invoked export control measures, directing Anthropic to bar foreign nationals—both abroad and domestically, including some of its own staff—from using Claude Fable 5 and Claude Mythos 5.
To fully adhere without selective enforcement challenges, the company implemented a total suspension for every user worldwide. Other Claude variants remain fully operational.
In its detailed public response, Anthropic expressed strong reservations about the rationale.
Company executives explained that the cited vulnerability involved a narrow prompting technique to examine code for minor, pre-existing software issues.
They stressed that similar outputs are readily available from competing models like OpenAI‘s offerings, without any special circumvention.
The firm characterized the government’s input as largely informal and pledged to publish further technical analysis within 24 hours to demonstrate the limited nature of the concern.
Anthropic underscored its proactive safety measures during development. Before launch, the models underwent extensive red-teaming involving government partners, independent experts, and internal reviews spanning thousands of hours.
Safeguards were tuned conservatively to minimize misuse risks in sensitive domains like cybersecurity, though this sometimes affects legitimate queries.
The company argued that a broad recall based on isolated, non-universal bypasses could impede progress across the entire sector if applied universally.
Market reaction was swift and negative. On secondary trading venues such as Hyperliquid, Anthropic-linked perpetual contracts declined approximately 3.7%, dropping to levels near $1,627 from peaks above $1,800 seen shortly after the models’ debut earlier in the week.
Although liquidity in these instruments is relatively thin, the drop signals investor caution regarding near-term valuation and the firm’s path to a public listing. This occurs as Anthropic, known for its safety-first philosophy, advances preparations for an IPO.
The company recently submitted confidential registration documents, positioning itself among several major AI players eyeing public debuts.
However, such regulatory actions introduce new variables that could influence investor sentiment and operational momentum.
Fable 5 and Mythos 5, unveiled just days prior, represented significant leaps in capabilities for software engineering, scientific reasoning, vision tasks, and long-context work.
Mythos 5, in particular, supported critical cybersecurity defense initiatives through programs like Project Glasswing.
Proponents now argue that temporary restrictions risk slowing beneficial applications in infrastructure protection and research, even as they prompt important debates on governance frameworks for powerful AI.
Anthropic indicated it is actively engaging to resolve the matter and restore availability promptly, viewing the directive as based on a misunderstanding. The resolution could ultimately impact / influence future interactions between regulators and innovators, potentially shaping how export rules and safety evaluations apply to evolving digital and AI enabled technologies.
