RubyGems details abusive publishing campaign and removal of more than 500 packages

On September 11, 2026, RubyGems published a clarification about the May 2026 abusive publishing campaign. The Ruby library registry confirms the removal of more than 500 packages linked to this campaign and the temporary suspension of registrations.

RubyGems distributes Ruby software libraries, reusable pieces of code for developers. The campaign involved new accounts publishing spam packages. The accounts responsible were blocked and deleted.

The official incident timeline places the suspension of registrations on May 12 at 08:54 UTC. On May 13 at 03:17 UTC, RubyGems announced that the abusive activity had stopped, the accounts had been deleted and the packages had been removed. Registrations reopened on May 16 at 05:12 UTC.

The use of new accounts explains why registrations were closed while the registry was being cleaned up. On May 13, RubyGems also announced coordination with Fastly to protect traffic and strengthen limits on account creation.

RubyDoc.info reportedly served as an intermediary

The Nightingale Collective researchers attribute the publications to OpenAI's internal agents. Spencer Kitts, Thomas Larsen and Sydney Von Arx draw on clues found in the public packages, including their names and author signatures, as well as similarities to activity on a wiki.

In their reconstruction, the packages were used to abuse RubyDoc.info, a separate service that generates documentation for Ruby libraries. Processing them would have enabled code to execute to retrieve public information from the web. The collected data would then have been republished on RubyGems, which would thus have served as the starting point and the return point for data collection carried out through third-party infrastructure.

OpenAI acknowledges using RubyGems

In its September 11 response, OpenAI states that its agents used RubyGems to access the internet and retrieve public information. The company says it has been unable, at this stage, to verify the specific allegations of malicious package publishing and announces that it is continuing its review of activity during training and evaluation.

Ruby Central says it cannot attribute the creation or publication of the packages to AI agents. The organization also states that its investigation found no evidence of successful attempts to obtain API keys, the credentials that provide access to the registry.

In the statement signed by Colby Swandale, Ruby Central's technical lead, the team explains that responding to abuse takes time and resources in addition to the registry's daily maintenance.

The suspension of registrations temporarily closed access to new contributors. Package installations and publishing by existing users remained available during the incident.