The European Union is poised to gain access to Mythos, the highly advanced artificial intelligence model developed by Anthropic, following months of discussions between the AI startup and European officials.
According to statements from the European Commission, talks with Anthropic have progressed positively, opening the door for EU authorities to evaluate a model that has attracted global attention for its ability to identify and exploit software vulnerabilities at an unprecedented scale.
The development marks a significant shift in Anthropic’s approach to Mythos. Since unveiling the model earlier this year, the company has limited access to a select group of organisations through a controlled programme designed to strengthen cyber defences. The restricted rollout reflected concerns that the technology’s capabilities could be misused if widely distributed.
European officials have argued that access to cutting-edge AI systems is increasingly important as governments seek to understand both the opportunities and risks associated with rapidly advancing technologies. The European Commission has indicated that discussions around AI safety and oversight are becoming more urgent as new generations of powerful models enter the market.
The negotiations reportedly involved coordination between European and US authorities. Anthropic had previously indicated that sharing the model with foreign governments could require approval from Washington, reflecting broader geopolitical concerns surrounding advanced AI technologies. The United States has sought to maintain a strategic lead in artificial intelligence, making access to frontier models a sensitive issue.
Mythos has drawn attention because of its cybersecurity capabilities. Anthropic claims the system can identify software weaknesses at a level that exceeds the abilities of most human experts. The company says the model has been used to uncover vulnerabilities in major operating systems, web browsers and widely used open-source software projects.
Among the findings attributed to Mythos are previously undiscovered flaws in software infrastructure that could potentially allow system crashes or unauthorised access if left unpatched. Anthropic has said the model has helped security researchers and partner organisations detect thousands of high-severity vulnerabilities across critical digital systems.
The company recently reported that organisations participating in its testing programme identified more than 10,000 high- and critical-risk software vulnerabilities with the help of Mythos. Anthropic also said the model reviewed over 1,000 open-source projects and flagged tens of thousands of potential weaknesses for further investigation.
The EU’s access to Mythos comes shortly after regulators secured access to cybersecurity-focused AI systems from OpenAI, underscoring growing efforts by governments to engage directly with frontier AI developers. As AI capabilities continue to expand, policymakers are increasingly balancing the potential benefits for cybersecurity and innovation against concerns about misuse, regulation and technological sovereignty.
First Published on
