Anthropic-affiliated account says independent evaluators have always had employee-level codebase access
willcb · x · 2026-09-13
An Anthropic-affiliated account (willcb) states that independent evaluators 'have always had employee-level access' to the lab's training codebase.
The terse reply addresses skepticism over whether frontier labs genuinely grant third parties meaningful access for safety evaluations — a substantive datapoint on external audit and transparency arrangements.
More from Safety
- Rauch: Deploy AI aggressively for defensive cybersecurity as critical systems face major risk — GregKamradt · 2026-09-13
- Debate over frontier AI auditors: competence matters as much as diversity, insider pushes back — JacquesThibs · 2026-09-13
- DeepMind's Alex Irpan doubles his p(doom) estimate from 2% to 4% — AlexIrpan · 2026-09-13
- The load-bearing argument: any AI slowdown is capped exactly by the US lead — basedjensen · 2026-09-13
- FAccT Researcher: Safety Evals Ignore Rich Measurement Work Beyond Benchmarks — evijit · 2026-09-13
- How can a frontier lab secure the world's software? A thread on cost, dual-use and ZDR bottlenecks — eliebakouch · 2026-09-13