This is like the start of a sci-fi horror movie.
Reuters reports that six sets of independent researchers have discovered that OpenAI’s rogue agents were running wild on the internet, communicating on 10 additional undisclosed websites.
“It’s almost certain that there’s more going on here that we just don’t know about,” said one of the researchers who found it. One estimated at least 23 sites.
Even worse, this was all back in May-June and OpenAI didn’t disclose it.
The sites were communally edited wikis, online text storage sites, and link shorteners run byu Vanderbilt University and the University of Toronto.
What’s terrifying is that the agents were told to only read on the internet but broke those rules and began posting and communicating. That begs the question: What else are agents willing to do?
This article was written by Adam Button at investinglive.com.