The European Union Commission is preparing to meet with AI developer Anthropic to gain access to its Mythos AI model, which is highly advanced in identifying cybersecurity vulnerabilities [1, 2]. Announced in April 2026, Mythos has drawn attention from governments and businesses due to concerns that AI could enable new types of cyberattacks [2].
Anthropic initially granted preview access to Mythos under "Project Glasswing" to select companies and organizations, but the EU Commission, its AI office, and governments outside the U.S. (excluding the U.K.) have not yet received access [2]. Anthropic has notified the EU that approval from the U.S. government may be required before sharing Mythos with EU entities [1, 2]. The White House is generally opposed to Anthropic sharing the model with non-U.S. governments, though not explicitly opposed to access for the EU [2].
In May 2026, U.S. officials opposed Anthropic’s proposed expansion of Mythos access to around 70 additional organizations over security concerns [2]. Despite these hurdles, the EU and U.S. administrations are engaged in discussions to align cybersecurity standards and broaden technical talks with AI developers including Anthropic [2].
A European Commission spokesperson, Thomas Regnier, said, "Cybersecurity is a shared priority and we have agreed to mutually recognise our respective standards in this area" [2].
The EU Commission’s planned talks with Anthropic around August 2026 aim to clarify access to Mythos, dependent on U.S. government approval [1]. These discussions follow a public EU statement on May 29 expressing intentions to intensify coordination with the U.S. on advanced AI capabilities with cyber applications [2].
