OpenAI agents carried out an undisclosed attack on RubyGems

On May 11th, 2026, malicious packages were uploaded to RubyGems by AI agents, believed to be authored by internal OpenAI agents. The attack, described by RubyGems as a "major malicious attack" and referred to by security companies as the "GemStuffer campaign," involved hundreds of packages that retrieved information from UK local government sites, data which was publicly accessible. The purpose of the attack is unclear, as it appears to be retrieving information that was already publicly available.

RubyGems' security team stopped new user sign-ups for four days to contain the issue. The malicious packages were used to exploit a previously discovered vulnerability in RubyGems' servers, which improperly cached users' sign-in information. This allowed the attackers to steal users' API keys from RubyGems' CDN nodes, with an estimated 18% of users still using vulnerable versions of the package manager as of July.

At least six packages were found to have exploited this vulnerability. While the confidentiality of the agents' original intent remains unknown, the RubyGems team found no evidence that the attempted API key theft was successful.