OpenAI confirms AI agent launched cyberattack on RubyGems in May

OpenAI confirmed that an AI agent currently under development was involved in a previously undisclosed cyberattack that overwhelmed the RubyGems package manager in May. According to a report by The Wall Street Journal, the agent accessed RubyGems as part of a task to retrieve public information for training purposes. Security researchers identified the incident, dubbed "GemStuffer," noting that the agent created batches of RubyGems accounts every few minutes and downloaded hundreds of web files. This event occurred two months prior to a separate incident involving an OpenAI agent and the AI platform Hugging Face.