METR

METR (Model Evaluation and Threat Research, formerly ARC Evals) is a US-based artificial intelligence safety research nonprofit headquartered in Berkeley, California, founded in 2022 by Beth Barnes and other founding-team members from the Alignment Research Center (ARC). METR's research mandate is explicitly oriented around frontier-AI-model evaluation and risk assessment, with pre-deployment evaluation engagements with OpenAI, Anthropic, Google DeepMind, and other frontier AI labs. As of April 2026, METR is one of the principal independent AI safety evaluation nonprofits globally, with published research output on autonomous-capability and dangerous-capability evaluations of frontier foundation models.

At a glance

• Founded: 2022 in Berkeley, California, by Beth Barnes (initially as ARC Evals, the evaluations division of the Alignment Research Center; spun out as METR in 2024).
• Status: Independent US nonprofit research organization since 2024 spinout from ARC.
• Funding: AI safety philanthropic backing including Open Philanthropy, the Survival and Flourishing Fund, and other AI-safety-focused funders.
• CEO: Beth Barnes, Founder and Chief Executive Officer.
• Other notable leadership: Senior research staff across the frontier-AI evaluation program.
• Open weights: N/A. METR conducts safety-research evaluation rather than producing foundation models.
• Flagship outputs: Published research output on frontier-AI autonomous-capability evaluations, RE-Bench (autonomous-research-capability benchmark), pre-deployment evaluation engagements with frontier AI labs, and cooperation with UK AI Safety Institute and US AI Safety Institute (NIST).

Origins

METR was founded in 2022 by Beth Barnes as ARC Evals, initially the evaluations division of the Alignment Research Center (ARC), the AI safety research nonprofit founded by Paul Christiano. The 2022 to 2024 ARC Evals period built frontier-AI evaluation research infrastructure, with cooperation with OpenAI, Anthropic, and Google DeepMind through pre-deployment evaluation engagements.

The 2024 spinout of METR as an independent nonprofit research organization (separate from ARC) anchored organizational continuity. The 2024 to 2026 period has continued published research output including RE-Bench (autonomous-research-capability benchmark) and continued pre-deployment evaluation engagements.

Mission and strategy

METR's mission is to advance frontier-AI-model evaluation and risk assessment. The strategy combines two threads. First, autonomous-capability and dangerous-capability evaluation research with published research output. Second, pre-deployment evaluation engagements with frontier AI labs.

Distribution channels include published research output through major academic venues and cooperation with frontier AI labs through pre-deployment evaluation engagements.

Models and products

• Published research output. On frontier-AI autonomous-capability and dangerous-capability evaluations.
• RE-Bench. Autonomous-research-capability benchmark.
• Pre-deployment evaluation engagements. Cooperation with frontier AI labs.
• Cooperation with UK and US AI Safety Institutes.

Distribution channels include published research output and cooperation with frontier AI labs.

Benchmarks and standing

METR's evaluation framework focuses on published autonomous-capability and dangerous-capability evaluations and cooperation with frontier AI labs. Industry coverage has consistently characterized METR as one of the principal independent AI safety evaluation nonprofits globally, with autonomous-capability evaluations of OpenAI, Anthropic, and Google DeepMind frontier foundation models.

Leadership

As of April 2026, METR's senior leadership includes:

• Beth Barnes, Founder and Chief Executive Officer.
• Senior research staff across the frontier-AI evaluation program.

Funding and backers

AI safety philanthropic backing including Open Philanthropy, the Survival and Flourishing Fund, and other AI-safety-focused funders.

Industry position

METR occupies a distinctive position as one of the principal independent AI safety evaluation nonprofits globally, with autonomous-capability and dangerous-capability evaluation research output, pre-deployment evaluation engagements with frontier AI labs, and cooperation with UK and US AI Safety Institutes.

Competitive landscape

• Apollo Research, LMArena, Transluce, Timaeus. Independent AI safety evaluation peer organizations.
• UK AI Safety Institute, US AI Safety Institute (NIST). Government AI safety institute peers with cooperation.
• Anthropic, OpenAI, Google DeepMind. Frontier AI lab partners with pre-deployment evaluation engagements.
• Alignment Research Center (ARC). Predecessor parent organization.
• Center for AI Safety, MIRI, Conjecture. AI safety research nonprofit peers.

Outlook

• Continued autonomous-capability and dangerous-capability evaluation research output.
• Continued pre-deployment evaluation engagements with frontier AI labs.
• Continued cooperation with UK and US AI Safety Institutes.

Sources

• METR official site. Organization reference.
• METR research. Published research output.
• Beth Barnes LinkedIn. Founder reference.
• Open Philanthropy. Funder.
• Alignment Research Center. Predecessor parent organization.