UK regulators review Anthropic AI model kept out of public release

Yesterday 09:40

By: Dakir Madiha

UK regulators review Anthropic AI model kept out of public release

Anthropic has unveiled a preview of its most advanced model, Claude Mythos, but opted not to release it publicly, prompting scrutiny from UK regulators assessing potential risks.

The model was introduced on April 7 as part of Project Glasswing, a cybersecurity initiative involving more than 40 major technology firms, including Amazon Web Services, Apple, Google, Microsoft, Nvidia, JPMorgan Chase, Broadcom, Cisco, CrowdStrike, Palo Alto Networks, and the Linux Foundation. Anthropic committed up to $100 million in usage credits and $4 million in direct funding to support open source security efforts.

Claude Mythos is currently restricted to a small group of partner organizations. According to reporting from TechCrunch, it is being used for defensive cybersecurity tasks, including identifying vulnerabilities in both proprietary and open source software. Logan Graham of Anthropic told PBS NewsHour that the model can perform complex tasks comparable to a full day of work by a human security researcher. He said the decision to limit access reflects concerns about misuse if deployed widely.

The announcement has generated strong industry interest. At the HumanX conference, coverage by CNBC described intense attention around Anthropic’s technology. Glean CEO Arind Jain linked the momentum to earlier success with Claude Code, which had already reached more than $2.5 billion in annual revenue by February.

Regulators in the United Kingdom have moved quickly to assess the implications. Reuters reported on April 12 that financial authorities are evaluating risks tied to the model’s capabilities. RTÉ noted that Anthropic’s decision to withhold public release could reshape expectations across the technology sector if the system performs as described.

Claude Mythos sits above Anthropic’s existing models, including Claude Sonnet 4.6 and Opus 4.6, both designed with large context windows. Pricing for restricted API access is set at $25 per million input tokens and $125 per million output tokens, significantly higher than earlier models.

Anthropic has not announced a timeline for broader release. Industry observers are divided between expectations of a near-term launch and the possibility of integration into a future Claude 5 system later in 2026.