OpenAI reportedly finds evidence that more of its agents ran amok

In Brief

Much has been made of the incidental successful which 1 of OpenAI’s agents broke retired of its sandboxed trial situation and proceeded to hack the AI hosting level Hugging Face. OpenAI has since launched an investigation into however the incidental occurred, which is inactive ongoing.

Now, anonymous sources person told Reuters that much of OpenAI’s agents are believed to person escaped their sandboxes. However, 1 root downplayed the severity, saying that with those escapes, the agents didn’t look to permission OpenAI’s web to hack into different company’s. TechCrunch reached retired to OpenAI for much information.

AI programs acting successful bizarre ways has seemingly go a weird astir bragging constituent for companies. The aforesaid week, Anthropic besides announced that it had discovered not one, but 3 instances successful which its agents had escaped trial environments and hacked different organizations.

AI companies have besides been accused of utilizing specified incidents for selling purposes — arsenic they make sizeable attraction and whitethorn underscore however almighty the companies’ products are. The flip broadside of that is that these disclosures are besides ramping up discussions of authorities regulations.

Subscribe for the industry’s biggest tech news