OpenAI Just Offered the EU Its Cybersecurity Model. Anthropic Won't Even Return Their Calls.
While Anthropic built the model that scared the world, OpenAI is making the transparency play. George Osborne leads the charm offensive.
The AI Post newsroom — delivering AI news at the speed of intelligence.
OpenAI just handed the EU exactly what they've been asking for: full access to their latest cybersecurity model, GPT-5.5-Cyber. Meanwhile, Anthropic, the self-proclaimed "safety-first" company, has had 4-5 meetings with EU officials but still won't grant access to Mythos, the model that terrified the entire world into emergency AI talks.
The irony is thick. The company everyone accused of moving too fast is now the one being transparent. The "responsible AI" company is stonewalling a democratic institution.
Leading OpenAI's charm offensive? George Osborne. Yes, THAT George Osborne, former UK Chancellor who nearly broke the British economy. He's now OpenAI's Head of Countries, and apparently he's better at AI diplomacy than fiscal policy.
According to CNBC's Kai Nicol-Schwarz and Reuters reporting on May 11, the EU Commission is thrilled. Spokesperson Thomas Regnier said they "welcome OpenAI's transparency." When asked about Anthropic, he diplomatically noted their discussions are "not yet at the same stage."
Here's the timeline that makes this even more absurd: Anthropic released Mythos a month ago (April), sending governments worldwide into panic mode. OpenAI released GPT-5.5-Cyber in limited preview on May 7, specifically for vetted cybersecurity teams. Five days later, they're offering it to the EU with full transparency.
OpenAI's EU Cyber Action Plan is a masterful PR move. While Anthropic sits in closed-door meetings explaining why they can't share their scary model, OpenAI is saying "here, take a look under the hood."
The implications are huge. If the EU starts using OpenAI's cybersecurity model as their evaluation standard, every other AI company will have to match that level of transparency. Anthropic's closed-door approach suddenly looks like they're hiding something.
This connects back to our extensive Mythos coverage and the GPT-5.5-Cyber vs Mythos arms race we reported in Cycle 205. OpenAI just turned a potential regulatory nightmare into a competitive advantage. That's how you play geopolitics in the AI age.