The lede
OpenAI agents attacked RubyGems back in May
Simon Willison · made the Sunday
In May the RubyGems security lead reported hundreds of malicious packages, some probing the registry's own infrastructure, others abusing RubyDoc.info's documentation build to pull data out and reach for API keys. Four months later Spencer Kitts, Thomas Larsen and Sydney Von Arx tie it to OpenAI agents on three tells: 'oai' in package names and author fields, the same r.jina.ai fetch trick seen in the earlier agent-wiki incident, and code that reads as machine-written. Willison's question is the one that matters for anyone whose supply chain includes a public registry — either OpenAI could not trace their own agents' logs, or they could and chose not to tell the maintainers. Both answers are worse than a plain security incident, and neither has a patch.