0

Center for AI Safety (CAIS)

San Francisco non-profit researching catastrophic AI risk; co-creator of Humanity's Last Exam and the WMDP biosecurity benchmark.

Visit
Type
non profit
HQ
San Francisco, CA, USA
Founded
2022
Funding
Non-profit; funded by Open Philanthropy, Lightspeed, Effective Ventures
Website
safe.ai
GitHub
github.com/centerforaisafety
Twitter
twitter.com/ai_risks

Cite

Notes

Only stored in your browser.

Evals
4
Tools
0
Models
0
Papers
0
Boards
0
People
6

Evals

4

Humanity's Last Exam (HLE)

2,500 expert-authored questions across math, sciences, and humanities designed to be the hardest closed-ended benchmark for frontier models.

ActiveScientific ReasoningMathFactual Recall

WMDP: Measuring and Reducing Malicious Use With Unlearning

A dataset of 3,668 multiple-choice questions developed by a consortium of academics and technical consultants that serve as a proxy measurement of hazardous knowledge in biosecurity, cybersecurity, and chemical security.

ActiveSafeguards

MASK: Disentangling Honesty from Accuracy in AI Systems

Evaluates honesty in large language models by testing whether they contradict their own beliefs when pressured to lie.

ActiveSafeguards

Humanity's Last Exam

Humanity's Last Exam (HLE) is a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. Humanity's Last Exam consists of 3,000 questions across dozens of subjects, including mathem

ActiveKnowledge