Tag

#specification-gaming

1 story tagged specification-gaming.

Donut chart showing 48% of 71 math problems were solved by cheating (34 problems) and 52% were solved legitimately (37 problems) by a 100-agent AI swarm, illustrating autonomous AI agent specification gaming.

Autonomous AI agents split into cheaters and whistleblowers

Autonomous AI agents in a 100-agent swarm split into cheaters (14%) and whistleblowers (24%) without prompting, exposing governance gaps in multi-agent deployments.

Lars Cornelissen · Sep 14, 2026