Independent benchmark comparing frontier AI safeguards through systematic adversarial testing and estimated costs of finding reusable jailbreaks.

Berkeley coworking hub providing workspace, research community, meeting facilities, and shared support for organizations and researchers working on trustworthy AI.

International workshop series bringing researchers, policymakers, and industry leaders together to discuss AI alignment research and build collaborations.

Living testing standard defining baseline defenses against known jailbreak techniques, with systematic composition methods for assessing frontier model safeguards.

Targeted research grants supporting academics and independent researchers developing AI safety solutions; currently considers expert-nominated researchers.

Weekly research presentations hosted at FAR.Labs and streamed remotely, sharing current work on trustworthy and secure AI.