METR

OrganizationHigh(4)

Model Evaluation and Threat Research

Resources

24

Citing pages

53

Tracked domains

1

Credibility Rating

4/5

High(4)

High quality. Established institution or organization with editorial oversight and accountability.

Tracked Domains

metr.org

Resources (24)

24 resources

		Authors	Summary
METR: Model Evaluation and Threat Research	web	—	S	34
METR's analysis of 12 companies	web	—	S	10
Measuring AI Ability to Complete Long Tasks - METR	web	—	S	8
METR: Common Elements of Frontier AI Safety Policies	web	—	S	7
Details about METR’s evaluation of OpenAI GPT-5	web	—	S	7
METR's Analysis of Frontier AI Safety Cases (FAISC)	web	—	S	4
METR Capability Evaluations Update: Claude Sonnet and OpenAI o1	web	—	S	4
Evaluation Methodology	web	—	S	3
METR’s GPT-4.5 pre-deployment evaluations	web	—	S	3
RE-Bench: Evaluating frontier AI R&D capabilities	web	—	S	2
The Rogue Replication Threat Model	web	—	S	2
Common Elements of Frontier AI Safety Policies (METR Analysis)	web	—	S	2
Research note: A simpler AI timelines model predicts 99% AI R&D automation in ~2032	web	—	S	1
Details about METR’s evaluation of OpenAI GPT-5.1-Codex-Max	web	—	S	1
METR's June 2025 evaluation	web	—	S	1
METR Publications	web	—	S	1
Evaluation methodology	web	—	S	1
MALT: A Dataset of Natural and Prompted Behaviors That Threaten Eval Integrity	web	—	S	1
METR's Autonomy Evaluation Resources (March 2024)	web	—	S	1
METR: Responsible Scaling Policies	web	—	S	1
METR (Model Evaluation & Threat Research) - About	web	—	S	1
2024 01 11 Dangerous Capability Evaluations	web	—	S	1
AI models can be dangerous before public deployment	web	—	S	0
AI models can be dangerous before public deployment (METR, November 13, 2024)	web	—	S	0

Citing Pages (53)

AI Accident Risk Cruxes AI Timelines Ajeya Cotra Alignment Research Center (ARC)Capability Elicitation AI Capability Threshold Model Autonomous Coding AI Governance Coordination Technologies Corporate AI Safety Responses Dangerous Capability Evaluations AI Safety Defense in Depth Model Emergent Capabilities Epistemic Virtue Evals Epoch AI Eval Saturation & The Evals Gap AI Evaluations Evals-Based Deployment Gates AI Evaluation Heavy Scaffolding / Agentic Systems International AI Safety Summit Series AI Safety Intervention Effectiveness Matrix AI Safety Intervention Portfolio Intervention Timing Windows AI Lab Safety Culture Frontier AI Labs (Overview)Large Language Models Long-Horizon Autonomous Tasks Mesa-Optimization Risk Analysis METR Third-Party Model Auditing Persuasion and Social Manipulation AI Proliferation AI Development Racing Dynamics Red Teaming Reward Hacking AI Risk Activation Timeline Model AI Risk Cascade Pathways Model Responsible Scaling Policies AI Safety Cases AI Safety Culture Equilibrium Model Sandboxing / Containment Scalable Eval Approaches Scheming & Deception Detection Self-Improvement and Recursive Enhancement Seoul Declaration on AI Safety Survival and Flourishing Fund (SFF)Situational Awareness Sycophancy Technical AI Safety Research Tool-Use Restrictions AI Risk Warning Signs Model AI Whistleblower Protections Why Alignment Might Be Easy

Publication ID: metr

METR | Publications | Longterm Wiki