Industry News 17 April 2026 3 min readResultsense via POLITICO Europe
EU AI Office locked out of Mythos as UK keeps edge
AI safety groups tell the European Commission its AI Office lacks Mythos access and the staff to evaluate it, while the UK AI Security Institute published technical analysis within a week.
TL;DR:
- The European Commission is not among the 40 organisations Anthropic has granted access to its Mythos model, a Commission spokesperson has confirmed.
- Eight AI safety groups have written to Brussels demanding that the EU AI Office’s safety unit quadruple to 160 staff by 2030.
- The UK’s AI Security Institute, by contrast, gained access and published a widely praised technical analysis within a week of Mythos’s announcement.
The EU’s AI Office has around 140 staffers, with 36 in the safety unit responsible for the most capable models. Critics interviewed by POLITICO say that is too few coders, too low in the Commission hierarchy, and too far from political leadership to respond to a Mythos-class release. The letter from the eight safety groups, first reported by POLITICO, asks the Commission to match the scale of its platform-enforcement teams.
The UK-EU gap
Several points of contrast are now visible. The UK AISI has access to Mythos and published a technical risk analysis. The Commission does not. UK Prime Minister Keir Starmer has an in-house AI adviser, Jade Leung, a former OpenAI lobbyist. The EU’s AI Office sits several management layers below Commission political operatives. Ciaran Martin, former head of the UK’s National Cyber Security Centre, told POLITICO that a non-regulatory state body — one without the power to fine — can often engage more effectively with AI labs than a regulator carrying a €35 million maximum penalty.
AISI’s proximity to government without regulatory teeth has become a quiet competitive advantage for the UK, even as it is politically vulnerable: the same arrangement that produces cooperation with Anthropic is the arrangement that an incoming government could dismantle in a weekend.
The hiring problem
The Commission is pushing to hire 38 additional AI Office staff as part of the AI law simplification package launched in November 2025, but those negotiations are unresolved. A further half-dozen safety, regulation and compliance hires are planned by the end of June. Salaries remain the bottleneck: the EU has to compete with industry wages and with London’s concentration of AI talent, a cluster that Anthropic’s same-week announcement of an 800-person London office will only deepen.
Looking forward
If Mythos becomes a regulatory template — vetted operator access, technical evaluations and phased release — then the authority with the best evaluators wins the sequencing. Today, that is AISI, not the AI Office. The short-term question is whether the Commission can credibly close the gap before the next frontier model lands. The longer-term question is whether Brussels can replicate a trust-based, non-regulatory relationship with labs while still holding the AI Act’s enforcement powers — the two roles that Ciaran Martin’s comment suggests pull in opposite directions.
Bankwatch 