safe.ai · Labs & Institutes
Nonprofit doing technical safety research, field building and compute support for researchers studying risks from advanced AI.
Center for AI Safety is one of 9 lab and institute tools on Intelsi. See all lab and institute tools or how we check facts.
// guides
What is AI safety? A clear 2026 explainer: the three kinds of AI risk, how safety differs from alignment and security, the main research areas and who does it.
Superintelligence vs AGI: AGI is roughly human level general intelligence; superintelligence far exceeds it. Lab definitions, DeepMind levels, and US SI.
EO 14434, signed September 29, 2026, tells federal agencies to say 'SI' instead of 'AI'. What it requires, what stays the same, and the November 28 deadline.
What does SI mean instead of AI? Under EO 14434, US agencies now say Super Intelligence. What SI covers legally and how it differs from superintelligence.
// related
METRmetr.orgNonprofit that evaluates frontier models for dangerous autonomous capabilities and measures how long a task AI agents can complete.Labs & InstitutesFreeHN 18
UK AI Security Instituteaisi.gov.ukUK government body that tests frontier models before and after release and funds research into AI security.Labs & InstitutesFreeHN 8
Redwood Researchredwoodresearch.orgNonprofit research lab working on AI control: techniques for safely using powerful models even if they are misaligned.Labs & InstitutesFreeHN 3
MIRIintelligence.orgThe Machine Intelligence Research Institute, one of the earliest organisations focused on risks from smarter than human AI.Labs & InstitutesFreeHN 3
Apollo Researchapolloresearch.aiAI safety organisation studying deceptive behaviour and scheming in advanced models through evaluations and interpretability.Labs & InstitutesFree
Google DeepMinddeepmind.googleGoogle's AI research lab behind Gemini, AlphaFold and many landmark results in reinforcement learning and science.Labs & InstitutesSee site