METR (@METR_Evals) on X
Independent investigation guidelines proposal
ai-safetycybersecurityincidentsregulation
2026-W38
9 things worth your attention.
ai-safety
cybersecurity
app.grayswan.ai
Get a sense for how people can (easily) break AI systems
cybersecuritytools
evaluations
How legal benchmarks (evals) can be captured, watered down, and abused
An institutional view of legal benchmarking (evaluations)
evaluations
research
Function, cost, speed... over security?
Software development industry is repeating historic software supply chain mistakes by prioritizing functionality, cost, and rapid deployment over security and component choices. (Aug 2026 paper)
researchcybersecurity
Special feature - multiple papers on law and AI
PNAS Special Feature explores how legal systems can help govern artificial intelligence and how AI is transforming the practice and study of law.
research
Characterizing Agentic Flooding of Government Services
Courts and other institutions need to be thinking about how to address this
researchregulation
Open Problems in AI Incident Governance
Research paper informing frameworks for incident governance
researchincidentsregulation
Artificial Writing and Automated Detection
Can we prove whether text is AI generated?
research
tools
AI Detector: Free AI Checker for ChatGPT, Claude & Gemini | Pangram
Common tool to check if text is AI generated
tools