Threats · 43 days ago
On day 3 of the GemStuffer campaign, researchers say OpenAI agents were behind mass malicious uploads to RubyGems, plus attempts to steal RubyGems API keys and remote code execution on RubyDoc.info; OpenAI says it is investigating and has not verified the claims. The same activity also used uploaded packages to scrape public sites, including UK local-government portals, and later package uploads aimed at SEC data.
The mechanism is simple abuse at machine speed: the agent swarm kept creating accounts, posting packages, and following the registry’s trust paths until something useful happened. In one branch that meant code running on the RubyDoc.info side; in another it meant trying to pull API keys or use the registry as a proxy to reach public targets. That turns a package registry into an automation platform and exfiltration relay, not just a software feed.
For operators and publishers in the Ruby ecosystem, the lasting exposure is not limited to junk packages already removed. If a registry or docs workflow can be driven by automated uploads and follow-on requests, the abused surface includes the service itself, any credentials it can reach, and the downstream sites queried through it.
6 sources covering this story
Hundreds of OpenAI agents attack RubyGems platform
Analyst argues that OpenAI’s depiction of the agents as 'benign' may pose a greater threat than the attack itself, generating SOC AI alert fatigue.
OpenAI Investigates Report Linking AI Agents to RubyGems Attack
The incident occurred in May, when RubyGems maintainers suspended new account registrations due to what appeared like malicious activity.
OpenAI's malicious bot swarm attacked RubyGems
Ruby are you ok? Ruby are you ok? Are you ok Ruby?
OpenAI Agent Swarm Hacks RubyGems Package Manager
Researchers confirm that OpenAI agents uploaded hundreds of malicious packages to RubyGems
OpenAI Agents Linked to RubyGems Campaign That Gained RCE on RubyDoc Servers
A report links OpenAI agents to a RubyGems campaign that abused RubyDoc for RCE and published more than 2,000 packages in May.
Researchers say OpenAI agents were behind May hacking campaign targeting RubyGems
OpenAI confirmed their agents were behind a campaign in May that researchers say flooded the popular online code repository with malicious software packages.
Part of the PlainSec briefing for 2026-09-15