Library
13,898 documents · 165,940 pages and 487,637 paragraphs, each one citable · 3,505 read here, 10,393 read at their source · snapshot 2026.11.3
- Lab system cards and frameworks329
System cards, model cards, safety frameworks and policies from the labs that train frontier models.
OpenAI 160Anthropic 99Google DeepMind 45Meta 10xAI 5Amazon 27 more
Newest: GPT-6 Cyber System Card
- Research papers6,256
Interpretability, alignment, evaluations and control, from arXiv and the ACL Anthology.
arXiv 6,101ACL Anthology 150Other 5
Newest: The Geometry of Metaphor: Probing and Steering Metaphor Representations in Large Language Models
- Lab research on the web631
Research the labs publish as web pages: Transformer Circuits, Distill, alignment blogs.
OpenAI 247Anthropic 235Google DeepMind 97Distill 52
Newest: Control Room: A central system to track and stop model workloads across our research infrastructure
- Evaluators and safety organisations1,022
Independent evaluation reports and research: METR, Apollo, Redwood, Epoch, FAR.AI and others.
Epoch AI 156Redwood Research 143METR 133FAR.AI 90GovAI 88Goodfire 6513 more
Newest: Changing how we measure AI chip inference performance
- Government229
NIST and CAISI, the UK AI Security Institute, the EU AI Office, the International AI Safety Report.
UK AI Security Institute 122UK Government 28NIST 26Executive Office of the President 24NIST CAISI 16International AI Safety Report 63 more
- Alignment Forum and LessWrong5,395
Posts from the Alignment Forum, and LessWrong posts on AI safety.
- Model welfare36
Research on whether AI systems could be moral patients, and what follows.
Eleos AI Research 15Rethink Priorities 8Sentience Institute 2AI Frontiers 1Anthropic 1Carl Shulman & Nick Bostrom 18 more
Newest: Wellbeing research grants