Evaluator Bench
Dario Amodei (Anthropic CEO), X, 2026-09-12; retrieved 2026-09-15 tier 3 self (anthropic)confirmed

Dario Amodei on X: 'We Must Pace the Frontier ... We'll provide third-party evaluators with permanent, employee-level access'

https://x.com/DarioAmodei/status/2098773920774074715

Announcement post (4:30 PM, Sep 12, 2026 per the page). fetch_text reads x.com status pages through their server-rendered title/OG text, so CI can check these spans (SHA-256 a9e72bbe0e27...). This is the 'employee-level' wording the paper uses; METR is not named in the post itself, only in the essay it links.

Signals citing this source

  • METR, A (for): Named by Anthropic as the example embedded evaluator: an ongoing, employee-like-access commitment with publication without editorial control; OpenAI's chief executive said OpenAI would do the same on access, without naming an evaluator or publication rights.

Ledger rows citing this URL

  • none