Skip to content
Longterm Wiki
Index
Division·Cybm2DQ6Zy·Record·Parent

Division: ARC Evals

Verdictunverifiable95%
1 check · 4/13/2026

The source text is entirely about METR (Model Evaluation & Threat Research) and its work on AI evaluations and safety policies. While the record claims 'ARC Evals' is a division/team within some organization with active status, the source document does not mention 'ARC Evals' at all. The source is about METR, not ARC (which typically refers to Anthropic Research Compute or similar). Without any mention of 'ARC Evals' in the source material, the claim cannot be verified or contradicted based on this source—it is simply not addressed.

Our claim

entire record
Name
ARC Evals
Division Type
team
Status
active
Notes
Evaluates frontier AI models for dangerous capabilities (e.g., autonomous replication). Spun out as METR in 2024 but ARC continues related eval work.

Source evidence

1 src · 1 check
unverifiable95%Haiku 4.5 · 4/13/2026

NoteThe source text is entirely about METR (Model Evaluation & Threat Research) and its work on AI evaluations and safety policies. While the record claims 'ARC Evals' is a division/team within some organization with active status, the source document does not mention 'ARC Evals' at all. The source is about METR, not ARC (which typically refers to Anthropic Research Compute or similar). Without any mention of 'ARC Evals' in the source material, the claim cannot be verified or contradicted based on this source—it is simply not addressed.

Case № Cybm2DQ6ZyFiled 4/13/2026Confidence 95%