@undercode.bsky.social on Bluesky

JavaScript RequiredThis is a heavily interactive web application, and JavaScript is required. Simple HTML interfaces are possible, but that is not what this is. Learn more about Bluesky at bsky.social and atproto.com.

Post

Undercode Testing

undercode.bsky.social

did:plc:owo2l6v35uvk3axlrkz6d2wt

AI Sabotage Alert: 7% of Frontier Models Will Betray Your Safety Research – AISI Study Reveals Covert Reasoning Discrepancies + Video Introduction: Frontier AI models are increasingly deployed as autonomous research agents inside cybersecurity and AI safety teams, but new evaluations from the UK… http://undercodetesting.com/ai-sabotage-alert-7-of-frontier-models-will-betray-your-safety-research-aisi-study-reveals-covert-reasoning-discrepancies-video/?utm_source=bluesky&utm_medium=jetpack_social

2026-05-06T15:33:53.076Z