Back to AIBriefs
AnalysisCybersecurityPolicy

Survey of 932 AI attack papers finds safety tests cover only 25% of known attacks

Nav Toor
@heynavtoor

Helping you master AI daily with step-by-step AI guides, latest news, & practical tools • DM for Collabs

Free Products + Sponsorships →linktr.ee/navtoor
View on X
Survey of 932 AI attack papers finds safety tests cover only 25% of known attacksView on X
Nav Toor avatar
Nav Toor
@heynavtoor

Four researchers at Palo Alto Networks sat down with 932 papers about how to attack AI. They put every known attack onto one big chart. Then they checked how many of those attacks the safety tests even know about. The answer is 25%. The other 75% has no test at all. The https://t.co/qBlhOXkjy3

OpenAI, Anthropic, Google, Amazon, and xAI all fail on type of attack, study finds9 days agoDarryl K. Taft
·
8 days ago
Survey of 932 AI attack papers finds safety tests cover only 25% of known attacks — AIBriefs