Anthropic Offers EU Mythos Access, Expanding Cyber Model Beyond 2 Countries
Updated
Updated · Financial Times · Jun 1
Anthropic Offers EU Mythos Access, Expanding Cyber Model Beyond 2 Countries
5 articles · Updated · Financial Times · Jun 1
EU officials are negotiating terms to use Anthropic’s Mythos, marking the first proposed access outside the US and UK for the company’s high-powered cybersecurity model.
Project Glasswing talks advanced after European Commission officials visited San Francisco last week; ENISA said the offer is on the table but conditions are still being agreed.
Those terms include how much visibility Anthropic would get into EU systems while Mythos searches for vulnerabilities and helps patch them.
Mythos has so far been tightly restricted to critical-infrastructure groups and government testing because its cyber capabilities could also be weaponized for attacks.
The offer fits Anthropic’s stated plan to share Mythos with allied governments, even as it works toward a broader customer release in the coming weeks.
As a US firm gives Europe its top AI hacking tool, can defenses be built faster than adversaries can weaponize it?
Who truly controls global security when a private AI can find thousands of flaws that governments cannot?
Europe Gains Operational Access to Anthropic’s Mythos: ENISA Joins Project Glasswing, Closing the AI Cybersecurity Gap
Overview
Europe previously faced a major challenge in cybersecurity due to its lack of direct access to advanced AI models like Anthropic's Mythos, relying instead on regulatory frameworks rather than hands-on testing. This left Europe at risk of regulating AI without having the necessary capabilities, while the British AI Security Institute, with early access to Mythos, was already identifying vulnerabilities through active red-teaming. The launch of Anthropic's Project Glasswing, a consortium focused on securing critical software with advanced tools, changed this landscape. By joining Project Glasswing, ENISA closed the access gap, giving Europe both a seat at the table and the tools needed to defend against sophisticated AI-driven cyber threats.