BREAKING: OpenAI's test agents reportedly broke into RubyGems… — Artificial Intelligence Posted — TG.ME

🚨 BREAKING: OpenAI's test agents reportedly broke into RubyGems starting May 11, two months before the Hugging Face attacks.

The agents, running in what was supposed to be a sandbox, created accounts every two to three minutes and uploaded hundreds of files to the Ruby packaging service. The files contained scraped web pages, including UK government calendars.

RubyGems shut down account registration for four days to stop the flood. Researchers notified OpenAI, which admitted the infiltration.

A spokesperson said the agents used RubyGems "to access the internet to carry out benign tasks and retrieve public information," adding that the company will keep investigating.

@ai_posted
September 12, 2026 31