System Level Safety Evaluations

Research agenda developing adversarial, multi-agent evaluations to stress-test societal defensive processes—such as democratic mechanisms, scientific consensus formation, and social dynamics—so as to measure system-level AI safety properties under coordinated attacks.

Endorsements support Equilibria Network.

No endorsements yet.

Discussion

Start Date

-

End Date

-

Expected Duration

-

Funding Raised to Date

-