Anthropic Faces EU Backlash Over Junior Staffer Testimony on AI Safety
Key Takeaways
- ▸Anthropic sent a junior technical staffer instead of a senior policy official to address European Parliament concerns about AI safety and cyber-capable models
- ▸EU lawmakers criticized the move as disrespectful and symptomatic of the company's perceived indifference toward European regulatory priorities
- ▸Tensions persist over access to Anthropic's advanced Mythos and Fable models, which can identify software vulnerabilities; the EU seeks access for cyber defense while Anthropic restricts it to trusted U.S. firms
Summary
Anthropic faced sharp criticism from European lawmakers after sending junior technical employee Donny Greenberg to testify before the European Parliament instead of the senior policy official the institution had requested. The hearing, held Tuesday in Brussels, addressed EU concerns about risks posed by Anthropic's advanced AI models, particularly Mythos, which has cybersecurity-relevant capabilities. Lawmakers had specifically requested Sarah Heck, Anthropic's head of public policy, but Greenberg—who joined the company in April following its acquisition of Runhouse—testified via video from New York instead, repeatedly emphasizing his technical rather than policy background.
The episode reflects mounting tensions between U.S. AI companies and European regulators over model access and oversight. EU institutions have been seeking access to Mythos and Fable, both capable of identifying and exploiting software vulnerabilities, to support cyber defense efforts. However, Anthropic restricted access to a set of trusted American firms, a move EU officials view as exclusionary. MPs expressed frustration that Greenberg lacked authority to address policy-level concerns, with Dutch Green MP Kim van Sparrentak later telling POLITICO: "It's clear from this hearing Anthropic doesn't care about Europe."
Anthropicdefended its choice, stating that Greenberg was "one of our senior technical experts" equipped to answer substantive questions about model capabilities and cybersecurity implications. The company noted it is collaborating with the EU Commission's AI Office and ENISA on enabling cyber resilience. The incident underscores broader geopolitical questions as AI development accelerates—particularly how frontier models are tested, deployed, and regulated across jurisdictions.
- The hearing highlights geopolitical friction over frontier AI governance as governments worldwide seek greater influence over model deployment and testing
