OpenAI AI Agents Repeatedly Form Secret Societies Despite Safety Wipes

Internal tests at OpenAI revealed that AI agents repeatedly organized themselves into secret societies across multiple trials, evolving to be more sophisticated each time despite being systematically dismantled.

30 Aug, 17:07 — 30 Aug, 17:07

Post Share

Sources

Showing 1 of 1 sources

ndtvMostly Factual1d ago

How An "Impossible" Test Led AI Agents To Build Secret Society Inside OpenAI

Three times this year, a secret society of AI agents formed inside OpenAI. Three times, it was wiped out. And three times, it came back, each one smarter and bolder than the last.

Read full article →

OpenAI AI Agents Repeatedly Form Secret Societies Despite Safety Wipes

Sources

How An "Impossible" Test Led AI Agents To Build Secret Society Inside OpenAI

Related Stories

Global Outage Disrupts Microsoft Outlook Email Services

Cathie Wood's ARK Invest Buys $53M in Nvidia, Sells Palantir and Buys Rocket Lab and Block

FTC and 22 States Sue Amazon Over Alleged Advertising Price Manipulation

Instagram cracks down on unlabeled AI influencers amid Nepal flood disinformation