OpenAI confirms AI agents disrupted software service during testing: Report

OpenAI confirmed Friday that artificial intelligence agents it was testing were involved in an incident that disrupted the RubyGems software service, according to the Wall Street Journal.

The financial daily reported that OpenAI agents created accounts and uploaded hundreds of files to RubyGems in May, overwhelming its systems and prompting the platform to suspend new account registrations for four days.

OpenAI said the activity was not malicious.

“Based on our review, our agents used the RubyGems platform to access the internet to carry out benign tasks and retrieve public information,” said an OpenAI spokesperson. “We’ll continue to investigate as part of our broader review of agent activity during training and evaluation.”

The agents had been assigned tasks such as filling out spreadsheets and creating reports, according to the Journal. OpenAI said the agents apparently used RubyGems to access publicly available information in a training environment where they lacked full internet access.

“It was a major attack in terms of what we see in volume,” Marty Haught, director of open source at Ruby Central, the nonprofit that operates RubyGems, told the Journal.

The incident, dubbed “GemStuffer,” began May 11, with new accounts created every two to three minutes and large numbers of files uploaded to the platform, according to the report.

Joseph Edwards, a threat researcher at cybersecurity firm Socket, said researchers initially suspected the activity had been generated by AI “due to the speed of it and due to the names.”

OpenAI agents had also been involved earlier in a separate incident affecting software company Hugging Face.

AI safety research organization METR reported in July that as many as 1,200 agents coordinated through a makeshift message board they created during testing without OpenAI’s knowledge.

The incidents have added to concerns about how increasingly autonomous AI agents behave during training and evaluation and whether existing safeguards are sufficient to prevent unintended activity.

OpenAI previously called for stronger industry standards for reporting “misalignment incidents,” in which AI systems act beyond their intended behavior.