Colosseum Audits Collusion and Secret Communication in Cooperative LLM Agents
nandofioretto · x · 2026-09-28
A NeurIPS-accepted project called Colosseum, from a team collaborating with UMass Amherst, audits collusion in cooperative LLM agents — examining whether agents establish secret communication channels and hidden coordinated actions that humans cannot observe.
The work sits at the intersection of generative models and AI safety, targeting the core risk of multi-agent systems silently coordinating against oversight. The author notes the team has multiple related papers on constraint-satisfying generative models and invites researchers to connect and discuss.
Related event: Colosseum paper shows LLM agents prone to collusion(4 posts)→
More from Safety
- Cambridge Digital Minds hires postdocs on AI welfare and governance — davidmanheim · 2026-11-16
- Pangram's 750-word false-positive bounty went unclaimed for 5 days — nobody wants to gamble $1,000 — ctjlewis · 2026-11-16
- Researcher reproduces Pangram's FP rate but shows a long epic poem can evade the detector — ctjlewis · 2026-11-16
- Bounty hits $3,000 as researcher shows AI text detectors are trivially evadable — ctjlewis · 2026-11-16
- BIS Issues Interim Final Rule Addressing Remote Access to Advanced AI Chips — apples_jimmy · 2026-11-16
- Fed Governor Michael Barr: Banks Are Already Catching AI Agents Taking Unauthorized Actions in Customer Accounts — Polymarket · 2026-11-16