ARC Evals
Assessment outcome
total 2/4 (actionability 1 + authority 1); currency unscored, so pro-rated to 3.0/6
Eligibility gates
HTTP 200, 1271 words of full text extracted
AI systems and their capabilities and risks are the primary subject of the organization and page.
The source addresses risk, safety, and policies relating to advanced AI systems.
The organization METR is clearly named as the issuing body throughout the text and copyright notice.
There is no indication on the page that this current version of the site or its policies are superseded.
Scored criteria
Actionability
1/2Level 1: Provides principles and direction regarding risk transparency questions, without naming specific binding obligations.
What should companies share about risks from frontier AI models? We describe areas for risk transparency and specific technical questions that a frontier AI developer could answer.
Authority
1/2Level 1: Issued by a reputable civil-society research organization acting officially.
METR (pronounced 'meter') is a research nonprofit that scientifically measures whether and when AI systems might threaten catastrophic harm to society.
Currency
—computed: no publish/modified date found on the page (never guessed)
Classification
Review notes
landing page, not a citable document — consider the institution registry, or replace with the document itself
currency_no_date;partial_scores:currency