The European Commission is currently in talks with OpenAI to gain access to a new AI model focused on cybersecurity. This could give Brussels an early look at a high-risk area in the AI field: tools that help security teams find and fix vulnerabilities.
OpenAI said on Thursday it would give preview access to its latest cyber model to selected cybersecurity teams. On Monday, the Commission said it welcomed OpenAI’s “proactive engagement,” including the company’s plan to grant access to the new model.
The bigger difference is with Anthropic. Anthropic has not yet given the EU access to Mythos, its well-known cybersecurity model.
The Commission’s public statements confirm this. Commission spokesperson Thomas Regnier said Brussels has had “four or five meetings” with Anthropic, but there have not been any talks about access to its AI models so far.
The Commission is drawing a line between two approaches
Regnier’s comments show that Brussels sees a real difference in how the two U.S. companies work with regulators. “With one (OpenAI), you have a company proactively offering to give access,” he said.
With Anthropic, the Commission has had “good exchanges” but is “not at a stage where we can speculate on potential access or not.” This shows the EU wants direct insight into models with serious cyber risks before they are widely used, not just to observe from afar.
This is important because AI models focused on cybersecurity are especially sensitive. They help defenders find weaknesses more quickly, but there are also worries that these tools could be misused.
CNBC describes the situation as OpenAI opening access while Anthropic holds back Mythos, showing that the debate about AI governance is now focused on access as much as innovation.
Why Mythos remains the sticking point
Anthropic’s model has attracted attention in Europe for weeks, but the company has not been as open as OpenAI. According to Global Banking & Finance, Anthropic has met several times with EU officials, but these talks have not led to discussions about access to its models. From the Commission’s point of view, this means Mythos is still more restricted than OpenAI’s latest cyber model.
This difference also shows that working with regulators is now part of the competition. OpenAI seems to use controlled access to show cooperation and build trust with policymakers. Anthropic is still talking with Brussels but has not given the same level of access to Mythos.
For the EU, this difference could be as important as how the models perform. Regulators need enough access to understand what these powerful AI systems can do.
A bigger test for AI oversight in Europe
Right now, the focus is on OpenAI, Anthropic, and their cyber-focused models. But the bigger question is whether governments can get real oversight of advanced AI before these tools become widespread.
Regnier’s comments suggest the Commission is rewarding companies that engage early and openly. CNBC’s report indicates that OpenAI is ahead in this area, while Anthropic is still more cautious with Mythos.
This leaves Europe facing a familiar but more urgent challenge: how to regulate advanced AI systems as their risks grow and companies decide how much to share with outsiders.
For now, OpenAI is getting closer to Brussels. Anthropic is still in talks, but has not yet agreed to give access.
