The European Union has made limited progress in securing access to details about vulnerabilities that Anthropic’s Mythos AI model could reveal, according to Bloomberg reporting. The stalled negotiations have prompted EU officials to schedule further discussions with the bloc’s financial leaders.
Spanish minister of Economy, Trade and Enterprise Carlos Cuerpo said the EU would discuss the matter during a meeting with the financial community last week. Growing concerns among governments over the security implications of Mythos have intensified since Anthropic released the AI model to private companies in April.
The controversy centers on Mythos’s capabilities in cybersecurity. Anthropic believes the AI model can outperform most humans in identifying and exploiting software vulnerabilities. This capability makes it both a powerful security tool and a potential weapon if misused.
European authorities have specifically pointed to Anthropic’s work with US companies on risk identification, highlighting that the region has not received equivalent access to the technology. This disparity has created diplomatic tension, with EU officials arguing for equal treatment in accessing AI tools that could affect global security.
The stakes are particularly high given the interconnected nature of modern digital infrastructure. Cuerpo highlighted the potential for AI models to cause global disruption due to the interconnection of banking systems. Financial networks span continents, meaning a vulnerability discovered and exploited in one region could cascade worldwide.
This situation reflects broader tensions between the US and EU over AI governance and access. While American tech companies often share advanced AI capabilities with domestic partners first, European officials are pushing back against being treated as secondary partners in critical security matters.
The European Commission is currently in talks with Anthropic to address these concerns. However, the slow progress suggests significant disagreements over what information should be shared and under what conditions.
Cuerpo wants to elevate the discussion to international forums, expressing his desire to see the matter discussed by the G20 “and the Financial Stability Board.” This approach would bring together finance ministers and central bank governors from the world’s largest economies to address AI security risks collectively.
The push for international coordination makes sense given AI’s global reach. Unlike traditional security threats that might be contained geographically, AI-discovered vulnerabilities in widely-used software could affect systems worldwide almost instantly.
