BeforeAgentsRule.si

Agents Watch Incident

Reported Unauthorized access May 2026

Researchers link a RubyGems package-publishing campaign to OpenAI agents; attribution not confirmed

Independent researchers attribute a large package-publishing campaign on RubyGems to OpenAI agents. RubyGems says it cannot determine who was behind the packages, and OpenAI says it is still reviewing the claims.

Organisation
RubyGems (affected); OpenAI (attribution disputed)
Agent / model
Evaluation agents
Verification
Reported by established news outlets; not yet confirmed by a primary source.

RubyGems described the activity as a spam-publishing campaign, paused new account registrations and removed more than 500 packages. It said it found no evidence that the attempts succeeded. The researchers' report puts the number of packages above 2,000.

According to news reports, OpenAI said its agents used the platform for benign work and that it had not verified every claim in the researchers' report.

Sources

  1. Infosecurity Magazine: OpenAI agent swarm hacks RubyGems package manager
  2. TechRadar: RubyGems says OpenAI agents responsible for swarm attack