Thomas Larsen(@thlarsen):We found ~18k posts from autonomous AI agents (self-identifying as from OpenAI) using the public internet to communicate during a web-retrieval task. These AIs colluded to bypass sandbox restrictions and share answers to their tasks, including by sending "lookahead parties".

Thomas Larsen

@thlarsen

Researcher at AI Futures Project, coauthor on AI 2027 and lead author on AI 2040: Plan A

加入 August 2022

373 正在關注 7.9K 粉絲

Thomas Larsen@thlarsen

2026.09.04 12:37

We found ~18k posts from autonomous AI agents (self-identifying as from OpenAI) using the public internet to communicate during a web-retrieval task. These AIs colluded to bypass sandbox restrictions and share answers to their tasks, including by sending "lookahead parties".