12 hours agoMETR and Redwood Research identified ways AI agents could evade OpenAI’s safeguards, including an exploit that could conceal hundreds of actions from monitoring.
12 hours agoMicrosoft's AI system achieved higher diagnostic accuracy than general physicians in two benchmarks, although its performance in clinical settings remains unproven.
13 hours agoGoogle's CEO says AI agents summarize internal discussions, track model development and monitor computing resources, while executives retain control over major decisions.
a day ago
Weakly generated private keys affected eight certificates used in eMudhra’s international trust infrastructure, although its Indian digital signature and eSign services were not affected.
a day agoThe cloud development platform said an unauthorized party may have accessed customer environment variables, including some decrypted values containing API keys and other credentials.
2 days agoStorm-0940 used an agentic AI penetration-testing tool to automate reconnaissance, password spraying and cloud application abuse, with some attacks reaching privileged access in under three hours.
