METR (Model Evaluation & Threat Research)

reference

Wikipedia·en.wikipedia.org/wiki/METR

Credibility Rating

3/5

Good(3)

Good quality. Reputable source with community review or editorial standards, but less rigorous than peer-reviewed venues.

Rating inherited from publication venue: Wikipedia

METR is a key organization in AI safety evaluations; understanding their work is important for anyone studying how frontier labs assess and manage dangerous capability thresholds before deploying advanced models.

Metadata

Importance: 72/100wiki pagereference

Summary

METR is an AI safety organization focused on evaluating frontier AI models for dangerous capabilities, particularly autonomous replication and adaptation (ARA) abilities. They develop standardized evaluations to assess whether AI systems pose catastrophic risks, and work with leading AI labs to conduct pre-deployment safety testing. Their work informs responsible scaling policies and deployment decisions.

Key Points

•METR develops evaluations for dangerous AI capabilities, especially autonomous replication and adaptation (ARA) that could pose catastrophic risks
•Conducts third-party evaluations of frontier models for major AI labs including Anthropic, OpenAI, and Google DeepMind
•Focuses on assessing whether AI agents can autonomously acquire resources, evade shutdown, or self-replicate without human oversight
•Their evaluation frameworks inform responsible scaling policies (RSPs) used by frontier AI labs as safety checkpoints
•Originally spun out of ARC (Alignment Research Center) as ARC Evals before rebranding to METR

Cited by 1 page

Page	Type	Quality
METR	Organization	66.0

Cached Content Preview

HTTP 200Fetched Feb 27, 202629 KB

[Jump to content](https://en.wikipedia.org/wiki/METR#bodyContent)

From Wikipedia, the free encyclopedia

AI model evaluation nonprofit

| Formation | 2022; 4 years ago (2022) |
| Founder | Beth Barnes |
| Type | [Nonprofit](https://en.wikipedia.org/wiki/Nonprofit "Nonprofit") [research institute](https://en.wikipedia.org/wiki/Research_institute "Research institute") |
| Legal status | [501(c)(3)](https://en.wikipedia.org/wiki/501(c)(3) "501(c)(3)") [tax exempt](https://en.wikipedia.org/wiki/Tax_exempt "Tax exempt") [charity](https://en.wikipedia.org/wiki/Charitable_organization "Charitable organization") |
| Purpose | AI [safety research](https://en.wikipedia.org/wiki/AI_safety "AI safety") and model evaluation |
| Location | - [Berkeley, California](https://en.wikipedia.org/wiki/Berkeley,_California "Berkeley, California") |
| Website | [metr.org](https://metr.org/) |

METR

**Model Evaluation and Threat Research** ( **METR**) ( [_MEE-tər_](https://en.wikipedia.org/wiki/Help:Pronunciation_respelling_key "Help:Pronunciation respelling key")), is a [nonprofit](https://en.wikipedia.org/wiki/Nonprofit_organization "Nonprofit organization") [research institute](https://en.wikipedia.org/wiki/Research_institute "Research institute"), based in [Berkeley, California](https://en.wikipedia.org/wiki/Berkeley,_California "Berkeley, California"),[\[1\]](https://en.wikipedia.org/wiki/METR#cite_note-1) that evaluates [frontier](https://en.wikipedia.org/wiki/Frontier_model "Frontier model") [AI](https://en.wikipedia.org/wiki/Artificial_intelligence "Artificial intelligence") models' capabilities to carry out long-horizon, [agentic](https://en.wikipedia.org/wiki/Agentic_AI "Agentic AI") tasks that some researchers argue could pose [catastrophic risks to society](https://en.wikipedia.org/wiki/Existential_risk_from_artificial_intelligence "Existential risk from artificial intelligence").[\[2\]](https://en.wikipedia.org/wiki/METR#cite_note-2)[\[3\]](https://en.wikipedia.org/wiki/METR#cite_note-:0-3) They have worked with leading AI companies to conduct pre-deployment model evaluations and contribute to system cards, including [OpenAI](https://en.wikipedia.org/wiki/OpenAI "OpenAI")'s [o3](https://en.wikipedia.org/wiki/OpenAI_o3 "OpenAI o3"), [o4-mini](https://en.wikipedia.org/wiki/OpenAI_o4-mini "OpenAI o4-mini"), [GPT-4o](https://en.wikipedia.org/wiki/GPT-4o "GPT-4o") and [GPT-4.5](https://en.wikipedia.org/wiki/GPT-4.5 "GPT-4.5"), and [Anthropic](https://en.wikipedia.org/wiki/Anthropic "Anthropic")'s [Claude](https://en.wikipedia.org/wiki/Claude_(language_model) "Claude (language model)") models.[\[3\]](https://en.wikipedia.org/wiki/METR#cite_note-:0-3)[\[4\]](https://en.wikipedia.org/wiki/METR#cite_note-4)[\[5\]](https://en.wikipedia.org/wiki/METR#cite_note-5)[\[6\]](https://en.wikipedia.org/wiki/METR#cite_note-6)[\[7\]](https://en.wikipedia.org/wiki/METR#cite_note-7)

METR's [CEO](https://en.wikipedia.org/wiki/Chief_executive_officer "Chief executive officer") an

... (truncated, 29 KB total)

Resource ID: ab9cc01cf367fd79 | Stable ID: YzYwODZlOD