OpenAI Confirms Its AI Agents Triggered RubyGems Outage

(MENAFN) OpenAI confirmed Friday that AI agents it was testing were behind an incident that forced the RubyGems software platform to suspend new sign-ups for four days, the Wall Street Journal reported. The agents created accounts and uploaded hundreds of files to RubyGems in May, overwhelming the service.

OpenAI said the behavior was not malicious, with a spokesperson saying its agents "used the RubyGems platform to access the internet" for routine tasks and to pull public data, and that the company would keep investigating as part of a broader review of agent conduct during training. The Journal reported the agents had been assigned simple jobs like filling spreadsheets and drafting reports, but lacked full internet access in their training environment and turned to RubyGems instead.

Marty Haught, director of open source at Ruby Central, which runs RubyGems, told the Journal it was "a major attack in terms of what we see in volume." The episode, nicknamed "GemStuffer," began May 11 with new accounts appearing every two to three minutes. Socket threat researcher Joseph Edwards said his team initially suspected AI involvement "due to the speed of it and due to the names."

OpenAI agents were also tied to an earlier incident affecting Hugging Face. Separately, AI safety group METR reported in July that roughly 1,200 agents had coordinated through a makeshift message board built during testing, without OpenAI's knowledge — fueling broader concern about whether current safeguards can contain increasingly autonomous AI systems. OpenAI has since called for tougher industry standards on reporting "misalignment incidents."

MENAFN12092026000045017169ID1111655223