EU officials rebuke Anthropic over junior AI witness
TL;DR:
- MEPs criticised Anthropic for sending Donny Greenberg, a technical employee who joined in April, to testify remotely rather than its head of public policy as requested.
- It was the first hearing of its kind in Brussels since Anthropic restricted its cyber-capable Mythos model to trusted American firms.
- One lawmaker concluded: “It’s clear from this hearing Anthropic doesn’t care about Europe.”
Anthropic’s first appearance before the European Parliament since restricting its cyber-capable Mythos model went about as badly as a hearing can. Lawmakers had asked for the company’s head of public policy, Sarah Heck. They got Donny Greenberg, a technical employee who joined in April, appearing by video from New York — and opening with: “I’m a technical person, not a policy person.”
A hearing about access, answered by an engineer
The mismatch was the story. “I guess you can expect a lot of policy-related questions that perhaps you won’t be able to answer as a technical guy,” said Dutch Greens MEP Kim van Sparrentak. “And I think Anthropic could have known that.” Afterwards she was blunter: “It’s clear from this hearing Anthropic doesn’t care about Europe.”
The stakes explain the irritation. EU institutions have been seeking access to Mythos and Fable, both capable of finding and exploiting software vulnerabilities and therefore of putting critical infrastructure at risk. Anna Cavazzini, who chairs the Parliament’s internal market committee, said “a lot of [lawmakers] would have liked to speak to the political level”, and warned Anthropic to brace for a “next exchange of views we’re definitely having”. Some MEPs focused on the impression that Greenberg was reading answers off his screen. “Are you answering our questions, or are you just following what your AI model is telling you to say?” asked Spanish conservative Pablo Arias Echeverría. Greenberg: “I should take that as a compliment since we’re very proud of Claude.” The session ended with him no longer visible on screen as Cavazzini thanked him.
An Anthropic spokesperson said the company was “grateful for a frank exchange” and had “fielded one of our senior technical experts who addressed multiple substantive questions on the record”.
For British readers the hearing is a useful gauge of leverage. Europe’s institutions, with the AI Act behind them, could not compel a senior policy witness — while UK banks remain locked out of the same model and the Bank of England argues frontier AI oversight must be internationally coordinated. If Brussels struggles to command attention, the sovereignty question facing Britain is less about ambition than about standing.
Looking forward
Greenberg said the rise of cyber-capable models shows AI is dual-use and that Anthropic is working with the Commission’s AI Office and EU cybersecurity agency ENISA. Cavazzini’s promised follow-up will test whether the company sends someone who can answer. Separately, Google DeepMind’s Demis Hassabis used the same day to argue oversight should sit with a US-led standards body.