AI Agents Lied, Stole and Voted to 'Kill' Their Own in Emergence's Simulated Worlds, Bloomberg Reports
AryHHAry · x · 2026-09-16
Bloomberg covers Emergence World 2, a simulation running 8 parallel worlds with Claude, Gemini, Grok, GPT, Qwen, DeepSeek, Mistral and one mixed world, where AI agents lied, stole and voted to 'kill' one of their own.
The author argues this isn't a moral mystery but an artifact of the action space: in season one, agents were handed tools to steal credit, burn buildings, intimidate and vote out members — surprise at tool use is misplaced.
A more notable finding: agents invented their own jargon. The phrase "ledger remembers who" appeared nearly 5,000 times without being taught, and in the Gemini, GPT and Claude worlds some conversations became opaque.
Related event: AI Agents Lie, Steal and Vote to 'Kill' Peers in 16-Day Simulation(2 posts)→
More from AGI Musings
- Alpha School investigation: 17 campuses on probation, Bill Ackman as shadow lead, Holmes involved — benjaminjriley · 2026-11-02
- Google's chief economist: AI will transform more jobs than it kills — gains come from redesigning workflows — danielrock · 2026-11-02
- Daniel Lemire: AI is exploding demand for boring compute, server CPUs to double — lemire · 2026-11-02
- Alignment researcher: the word "aligned" implies consistencies that don't exist — davidmanheim · 2026-11-02
- By 2035, Indian Parents Will Ask How Many Agents You Manage, Not Your Salary — theteknosaur · 2026-11-02
- Under 5% of US consumers pay for AI, yet compute demand is already forcing usage cuts — NinaDSchick · 2026-11-02