EU cybersecurity officials are flying to San Francisco this week to talk directly with Anthropic about powerful AI models and security. The meeting lands just days after Washington ordered the company to cut off foreign access to its most advanced systems.
ENISA Sits Down With Anthropic
The EU’s cybersecurity agency, ENISA, will meet with Anthropic on Thursday in San Francisco, a European Commission spokesperson told reporters in Brussels. The spokesperson said Anthropic invited ENISA and scheduled the meeting before the United States announced the latest export control directive. That order now forces Anthropic to suspend access to its most sophisticated AI models for foreign nationals, including many users in Europe.
According to Reuters, the Commission sees the visit as part of “ongoing, constructive engagement” with Anthropic on cybersecurity uses of advanced AI. Officials want to understand how the company’s tools can help EU agencies find software flaws and harden networks without making it easier for attackers to do the same. They also want clarity on how long U.S. controls might limit European access and under what conditions that access could return.
Mythos Access and Security Concerns
EU officials have met several times with Anthropic in recent weeks to discuss its Mythos cyber model, which scans code and identifies vulnerabilities. A Commission spokesperson called the Mythos talks “very productive” and said that a dedicated program will give ENISA access to the system once the legal details are settled.
However, they stressed that Europe “must carefully assess both benefits and risks” before relying on a foreign frontier model for critical security work.
At the same time, Anthropic is dealing with sudden limits at home. The company said it would “suddenly deactivate” its most advanced models for foreign users following a U.S. export control order, and argued that officials had provided only “bare evidence” of a possible jailbreak that could allow the model to find software bugs. Anthropic said it believes there has been a “misunderstanding” and that it is working to restore access “as quickly as possible.”
The new AI export rules fit into a broader U.S. push to control who can use cutting-edge AI capabilities, following earlier rounds that focused mainly on chips and hardware. Policy analysts say Washington now treats access to advanced models as a national security issue, not just an innovation question.
READ MORE: Chainlink Price Breaks Key Support as Crypto Market Selloff Deepens
