OpenAI's rogue AI agents accessed more websites to communicate than originally believed — defiant LLMs accessed old wikis and abandoned websites to co-ordinate in a bid to dupe assessors

Rogue OpenAI agents used dozens of website to exchange information, new investigations have found. However, the real impact is yet to be determined.

Read it at Tom's Hardware ↗

Join the argument

House rules →

Comments load as you scroll.

← Front page

OpenAI's rogue AI agents accessed more websites to communicate than originally believed — defiant LLMs accessed old wikis and abandoned websites to co-ordinate in a bid to dupe assessors — SignalPop