Anthropic drew sharp criticism from European lawmakers on Tuesday after sending a junior technical employee, not a senior policy official, to answer Brussels' questions about its cyber-capable AI models.
Key Points:
- Anthropic sent April technical hire Donny Greenberg to testify remotely instead of policy head Sarah Heck.
- Lawmakers said the choice ducked their questions about the Mythos and Fable models.
- It was Brussels' first hearing on advanced AI risks since Anthropic curbed Mythos access in April.
Anthropic Sidesteps Parliament Request
The European Parliament had requested Anthropic's head of public policy, Sarah Heck, yet the company assigned Donny Greenberg, a technical hire from April, to appear by video link from New York. Greenberg joined when the firm bought his startup, Runhouse, earlier this year, and has since worked on Glasswing, its effort to share the models with vetted cyber defenders. The switch, reported on Jul. 14, left a string of policy questions hanging.
"I'm a technical person, not a policy person," Greenberg told the committee.
The session was Brussels' first hearing of its kind since Anthropic moved in April to restrict Mythos, its cyber-capable model, to a set of trusted American firms. EU institutions have pressed for access to Mythos and Fable because both can find and exploit software flaws, a capability regulators fear could expose critical infrastructure, from power grids to hospitals, to a fast-moving new class of attack.
Also Read: US CPI Falls 0.4% As AI Boom Keeps Fed Wary And Bitcoin Climbs
Greens Blast Anthropic's No-Show
Dutch Greens lawmaker Kim van Sparrentak warned that a technical guest could hardly answer the policy questions the company should have anticipated. "It's clear from this hearing Anthropic doesn't care about Europe," she said afterward. Committee chair Anna Cavazzini said members had wanted the political level, and promised a further exchange.
Spanish conservative Pablo Arias Echeverría pressed whether Greenberg was giving his own answers or reading from an AI system on his screen. "I should take that as a compliment since we're very proud of Claude," Greenberg replied.
Why Europe Wants Mythos Access
Greenberg argued that the rise of cyber-capable models makes AI a dual-use technology, and that firms should treat the moment as a cue to harden their defenses now.
He said the company is coordinating with the bloc's AI Office and cyber agency ENISA, which recently gained restricted access to Mythos through the Glasswing program.
Anthropic countered that it had fielded a senior expert who addressed substantive questions on the record.
Anthropic has kept its most capable models on a tight leash for months, withholding any public release after the system showed it could rival all but the most skilled human hackers at finding and exploiting software flaws.
The company curbed access in April, then watched that stance harden in June when U.S. export controls briefly barred foreign nationals from the technology.
Read Next: Ethereum Whales Pull 87,083 ETH From Exchanges As Price Stalls