Skip to content
anthropic.com·

🔒Anthropic Onboards METR Team for Independent AI Assessments

Anthropic opens its doors to independent AI evaluators

TL;DR

Anthropic has committed to independent, third-party assessments of its AI models and safety practices, starting with METR evaluators gaining employee-like access. This marks a significant step in transparent AI development.

Anthropic has committed to independent, third-party assessments of its AI models and safety practices, starting with METR evaluators gaining employee-like access. This marks a significant step in transparent AI development, ensuring that evaluations are thorough and unbiased. The move is part of Anthropic's pledge to the White House Accord on Super Intelligence, aiming to set a new standard for AI safety and ethics. METR will publish detailed risk assessments, and Anthropic has onboarded six METR employees with full access to internal tools and systems, starting in early discussions with other organizations like Hugging Face.

Anthropic Onboards METR Team for Independent AI Assessments — anthropic.com

Key Points

1

Anthropic signed the White House Accord on Super Intelligence, pledging to partner with independent external auditors or evaluators.

2

METR will publish detailed risk assessments, with Anthropic providing full access to internal tools and systems.

3

Anthropic is working with SecureBio and Accenture as evaluators, with SecureBio reviewing the chemical and biological weapons chapter of Anthropic's August Risk Report.

4

Anthropic is in early discussions with several organizations, including Hugging Face, to expand its independent evaluation program.

5

Anthropic is developing three tiers of access for evaluators, with SecureBio currently in the first tier.

Why It Matters

If you're developing or using AI models, Anthropic's commitment to independent assessments sets a new standard for transparency and safety. METR's detailed risk assessments will provide valuable insights into the risks and capabilities of frontier AI models, affecting how teams approach AI development and deployment.

AnthropicMETRAI EthicsWhite House AccordRisk Assessments

Comments

Subscribe to join the conversation...

Be the first to comment

Enjoyed this article?

Get it daily. 7am. Free. Reads in 5 minutes.

Join 3,569 builders reading daily.

Also get