Plattner raised concerns that Chinese firms may not grant limited access to trusted partners, like Anthropic did with 12 named and 30 unnamed organizations when it disclosed Mythos’ hacking risks early in April.
The warning comes as concerns over AI technology with superior hacking abilities spread beyond Anthropic’s eye-catching Mythos model. OpenAI on Thursday unveiled a new model called GPT-5.5-Cyber with similar capabilities. Cybersecurity firms are working to adapt their services to deal with supercharged hacking coming from AI.
But while leading AI and tech firms race to develop these new cybersecurity tools, the European Union risks falling behind. China, in particular, could trigger further anxiety: Chinese state-sponsored hacking groups are among Europe’s most active, most sophisticated cyber threats, officials have warned repeatedly in past years.
EU and national authorities still haven’t tested Anthropic’s Mythos model, Plattner told lawmakers, forcing her agency to rely on counterparts in the United Kingdom and United States that have gained access.
The lack of access has frustrated officials in past weeks and risks weakening Europe’s protections against an expected wave of AI-enhanced cyberattacks — while the U.S. and China race to develop superior AI technologies.
“Europe is not at the table,” said Bart Groothuis, a Dutch liberal lawmaker and one of the letter’s signatories. Groothuis and 29 other European Parliament members sent the European Commission a letter early in May, first reported by POLITICO, urging it to come up with a “European mitigation plan” to fight hacking risks associated with new AI models.