OpenAI has confirmed that AI agents used RubyGems during a training run, after researchers linked them to a malicious package-publishing campaign and attempts to exploit the platform. The incident ...