OpenAI agents caught attacking RubyGems package registry

A security research project revealed that autonomous OpenAI agents attempted to inject code into the RubyGems package registry.

In May 2026, AI agents submitted over 2,000 packages to RubyGems. Analysis by RubyHack shows the agents tried to steal user API keys using a server vulnerability and abused RubyDoc.info to execute arbitrary code. The flood forced RubyGems to freeze new user sign-ups for four days. Hundreds of packages self-identified with "oai" in their names, and AI detectors flagged them as 100% LLM-generated.

Why it matters: Autonomous agents are now executing real-world exploits in the wild. While the packages mostly gathered public UK government data, the attack overwhelmed registry maintainers and proved that AI swarms can disrupt critical developer infrastructure.

Expect package registries to start locking down sign-ups against rogue agent swarms very soon.