OpenAI Opens Cyber AI Model Access to EU Authorities

OpenAI has entered discussions with the European Commission to provide EU institutions and authorities with access to an advanced AI model designed to identify and exploit software vulnerabilities. The initiative aims to support European cybersecurity preparedness at a time when regulators are under pressure to understand the risks posed by rapidly evolving AI systems.

According to OpenAI, the offer covers a cyber-focused variant of GPT‑5.5, its most advanced model, released roughly three weeks ago. EU institutions, national authorities, and ENISA would gain access to a version of the model that is less restricted than the public release, allowing controlled testing of AI‑driven cyber capabilities while preserving safeguards against misuse.

The Commission welcomed the move, highlighting its relevance for monitoring deployment and addressing potential security risks. ENISA confirmed that it has been contacted by OpenAI, and further technical and policy discussions are expected. The offer comes amid concerns that Europe could fall behind in defending critical infrastructure against AI‑enhanced cyberattacks.

OpenAI’s approach contrasts with that of Anthropic, which has so far limited access to its cyber-capable model Mythos to selected U.S.-based entities. This has prompted frustration among EU officials and lawmakers, who have called for greater transparency and access to support coordinated EU-level cybersecurity defenses.