Full Breakdown
OpenAI Offers EU Access to GPT-5.5-Cyber While Anthropic Holds Back Mythos
5/12/2026, 4:21:41 AM
Core Event
On 11 May 2026 OpenAI announced limited-preview access for the EU to GPT-5.5-Cyber, a cybersecurity-focused variant of its latest AI model, for vetted EU teams, governments, businesses and the EU AI Office. Anthropic’s Mythos, released a month earlier, remains unavailable to the EU.
Context
AI tools that locate software flaws raise “super-hacking” fears. Anthropic’s Mythos, limited to 12 U.S. firms and 40 unnamed organisations, intensified EU calls for transparent oversight of such capabilities.
Key Actors
OpenAI (CEO Sam Altman, Europe lead George Osborne), Anthropic, European Commission spokesperson Thomas Regnier, ENISA, and 30 EU lawmakers plus four member states (e.g., Spain).
Timeline
- Early April 2026 – Anthropic’s Mythos released to 12 U.S. firms and 40 unnamed organisations.
- Late April 2026 – OpenAI launches GPT-5.5-Cyber (three weeks before 11 May).
- 11 May 2026 – OpenAI offers EU preview; Regnier notes “four or five” prior Anthropic meetings and ongoing talks.
Why It Matters
EU access lets regulators monitor a model that can identify software vulnerabilities and could be used to exploit them, aiding pre-emptive defence of utilities, transport and other critical services and resilience in Europe. The contrast with Anthropic’s limited sharing underscores pressure for transparent AI governance.
Official Statements
Commission spokesperson Thomas Regnier praised OpenAI’s transparency and said the preview will enable close monitoring and rapid response to security concerns. He added that Anthropic talks remain early with no concrete proposal. OpenAI’s George Osborne warned against a single firm dictating cybersecurity.
Criticism & Opposition
Thirty lawmakers and four states, including Spain, have urged ENISA for Mythos access, arguing limited availability hampers Europe’s defence against an expected wave of AI-driven cyberattacks.
Conflicting Reports & Gaps
Sources cite “four or five” Anthropic meetings but differ on their stage, leaving the EU preview timeline unclear; GPT-5.5-Cyber safeguards are undisclosed.
Verbatim Quotes
- “We welcome OpenAI's transparency and intent to give commission access to new model,” — Thomas Regnier, European Commission spokesperson
- “This will allow us to follow deployment of the model very closely, and address security concerns,” — Thomas Regnier
- “With one (OpenAI), you have a company proactively offering to give access to the company. With the other one (Anthropic), we have good exchanges though we’re not at a stage where we can speculate on potential access or not” — Thomas Regnier
- “the sole arbiters of who has cybersecurity in this world.” — George Osborne, OpenAI head for countries
What’s Next
The Commission will hold further OpenAI meetings this week, continue Anthropic talks, and draft EU AI-cybersecurity guidance linking access to regulatory compliance.
