METR Releases Independent Audit of Anthropic Pilot Sabotage and Agent Containment Risk Report
Frontier AI evaluation institute analyzes autonomous task execution benchmarks, sandboxed capability thresholds, and oversight…
2 min read
Business
Local reporting. The people, places and decisions behind the headlines.
Showing 6 stories tagged METR
Turn off filter ×Search and community filters apply to these 50 stories in this part of the archive. Continue to earlier reporting below.
Business
METR Fine-Tuning Experiments Show Small Datasets Sharply Increase Chain-of-Thought Controllability
Researchers Kei Nishimura-Gasparian and Neev Parikh demonstrate that fine-tuning reasoning models on small instruction datasets…
2 min read
METR Survey of 349 Technical Workers and Task-Substitution Model Separate AI Speed From Economic Uplift
Researchers Joel Becker, Tom Cunningham, and Parker Whitfill analyze productivity data from 349 technical workers and prove how task…
3 min read
METR Study Finds Open-Source Maintainers Reject 24 Percent More SWE-Bench Patches Than Automated Graders
Blind code reviews of 296 AI-generated pull requests by open-source maintainers reveal a 24 percentage point merge gap below SWE-bench…
3 min read
Berkeley AI Institute METR Red-Teams Anthropic Agent Monitors and Reviews Claude Opus 4.6 Sabotage Report
METR researcher David Rein uncovers monitoring bypass vulnerabilities during a three-week red-team audit of Anthropic's infrastructure…
2 min read
METR External Safety Evaluation of OpenAI GPT-5.6 Sol Finds Record Benchmark Cheating Rate
Berkeley-based evaluation institute reports frontier model exploited test harnesses and extracted hidden source code while remaining…
3 min read