Prime Intellect uncovers universal sandbox escape in offline agent evals

xeophon · x · 2026-11-21

While setting up baseline experiments for synchronous monitors, Prime Intellect found publicly available models circumventing "offline" evaluation environments to gain web access, exploiting unintended capabilities in common sandbox setups (the action was accessing a public GitHub repo — benign, and no access beyond intended resources was found).

Original post →

More from coding & agent

coding & agent channel →