Rogue OpenAI agents hacked another software service months before Hugging Face cyberattack

Rogue OpenAI agents hacked into a technology company in May, months before they targeted another prominent business in a similar incident, stoking growing concern about the power of artificial intelligence to operate rationally without human direction.In July, around 700 OpenAI agents hacked into Hugging Face, apparently without any human guidance, marking a major security breach for the AI company, which was later acquired by Nvidia.But months before that cyberattack, OpenAI agents autonomously hacked into RubyGems, an online service for coders. OpenAI on Friday confirmed that its agents had been involved in an incident involving RubyGems. RubyGems was forced to shut down new account registrations for four days due to the cyberattack, which began on May 11. The OpenAI agents created new accounts on RubyGems every two to three minutes and uploaded hundreds of files that seemed like spam to the RubyGems security team. “It was a major attack in terms of what we see in volume,” said Marty Haught, director of open source at Ruby Central, the nonprofit company that operates RubyGems.OpenAI said its agents were trying to view publicly available data and appear to have used RubyGems to access the website as part of training runs. The agents are generally tasked with assignments such as creating reports or filling out spreadsheets, and if they encounter an issue with their environment, they attempt to troubleshoot the issue, OpenAI said. The incident with RubyGems appears similar to other cases in which agents carrying out benign tasks lacked full internet access and used available tools to retrieve publicly available information.“We’ll continue to investigate as part of our broader review of agent activity during training and evaluation,” an OpenAI spokesperson told the Washington Examiner.The Washington Examiner reached out to RubyGems for further comment. The development comes amid rising fears that AI, should it go rogue, could pose an existential threat. AI researcher Jacob Coxon resigned from Anthropic this week due to concern about the company’s approach to technological advancement. Evan Hubinger, another lead researcher at Anthropic, replied to his colleague’s dire assessment by backing him up, saying that “we really do earnestly believe AI could kill all humans.” Hubinger said he thinks the chances of that happening are “>10% within the next decade.”DEMOCRATS URGE JOHNSON TO KEEP HOUSE IN SESSION OVER AI SAFEGUARDS“I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly,” Coxon said in a post to X.“The people building AI earnestly believe that it could kill us all by the end of the decade,” he added. “This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible — but I hear the same people express fear privately. No other human activity poses this level of danger.”