# Anthropic gives METR employee-level access for safety reviews

> Anthropic lets six external evaluators access some AI systems to publish safety findings without company veto, while keeping customer data private.

- Type: AI news
- Page: https://completeaitraining.com/news/anthropic-gives-metr-employee-level-access-for-safety/
- Published: 2026-10-16
- Updated: 2026-10-16
- Author: Jeroen Erne
- Jobs: Executives and strategy, Government, Legal, PR and communications
- Reading time: 1 min

Anthropic has granted six external evaluators from METR employee-level access to selected AI systems for safety reviews. The arrangement allows these evaluators to publish risk findings without the company holding a veto over the content of their reports.

The access model is designed to strengthen independent assurance by giving testers a closer look at system internals than typical external audits permit. It explicitly excludes customer data and information protected by legal privilege, preserving standard confidentiality boundaries.

The move sets a new procedural benchmark for third-party oversight in the AI sector, placing Anthropic ahead of voluntary transparency norms while stopping short of full technical disclosure.

---

This is the Markdown form of https://completeaitraining.com/news/anthropic-gives-metr-employee-level-access-for-safety/. The whole site is mapped at https://completeaitraining.com/llms.txt.