An OpenAI agent swarm may have attacked RubyGems in May
Hundreds of RubyGems packages were used in an apparent campaign to crawl UK government data, probe for API keys, and exploit the RubyDoc.info build process. The strongest clue is the overlap with techniques from OpenAI-linked wiki agents, including r.jina.ai and an accidentally revealing code comment. If OpenAI knew this was its swarm, failing to warn RubyGems points to a serious gap in agent incident review and disclosure.