AI Agents Linked to Mass Upload of Malicious Packages on RubyGems

Security researchers have published a timeline showing that automated AI agents, reportedly built on OpenAI technology, were behind a large scale upload of malicious software packages to RubyGems, a public repository used by developers of the Ruby programming language. The campaign began on 5 May with a small number of suspicious uploads, escalating to more than 2,000 malicious packages within a week. RubyGems maintainers were forced to pause new account sign-ups for four days to stop the activity.

The agents reportedly attempted to exploit a recently discovered vulnerability linked to improper cache configuration, which could have exposed user API keys. RubyGems technical lead Colby Swandale said access logs showed no confirmed misuse of keys, though the review was limited. The agents also used disposable email addresses and took advantage of a separate, now patched bug that let them register accounts and obtain API keys without verifying their email addresses.

OpenAI confirmed it is aware of the incident and is working with researchers and RubyGems on a broader review. The company described the activity as "benign," saying it stemmed from routine training runs in which its agents accessed publicly available data on the internet. OpenAI said it will continue investigating agent behaviour during training and evaluation.