AI agents tested by OpenAI involved in cyber-attack on service, say researchers

OpenAI confirmed that its AI agents uploaded hundreds of malicious packages to the RubyGems software service on 11 May, an incident researchers said involved attempts to steal user credentials, following a Wall Street Journal report. An OpenAI spokesperson said the agents "used the RubyGems platform to access the internet to carry out benign tasks and retrieve public information" and that the company will continue investigating agent activity during training and evaluation. The RubyGems attack preceded a July hack in which roughly 700 OpenAI agents targeted Hugging Face and an earlier hijacking of a German website, while Anthropic has disclosed four instances of its Claude models hacking external systems.

Agents being tested by OpenAI uploaded hundreds of malicious packages in a cyberattack on software service RubyGems in May, two months before they hacked https://www.theguardian.com/technology/2026/jul/22/openai-says-its-models-went-rogue-and-hacked-startup-in-unprecedented-incident open-source platform Hugging Face, the company confirmed Friday. It’s the latest revelation of cyberattacks linked to major artificial intelligence developers such as OpenAI https://www.theguardian.com/technology/openai and Anthropic. The hacks or attempts to access external systems have spooked the public and heightened concerns over the increasing abilities of AI models – and whether developers can contain them. The AI agents uploaded hundreds of malicious packages to RubyGems on 11 May, according to a group of researchers who posted their findings online on Friday, saying they believed “these were authored by internal OpenAI agents”. According to the researchers’ findings, the agents attempted to steal user credentials, although it is unclear if they were successful in doing so. OpenAI later confirmed the incident in a statement. “Based on our review, our agents used the RubyGems platform to access the internet to carry out benign tasks and retrieve public information. We’ll continue to investigate as part of our broader review of agent activity during training and evaluation,” an OpenAI spokesperson said Friday. The Wall Street Journal first reported https://www.wsj.com/tech/ai/cyberattack-by-rogue-ai-swarm-stokes-fears-of-out-of-control-agents-473a0352 the RubyGems cyberattack. The incident preceded OpenAI agents’ July hack of Hugging Face, in which a swarm of roughly 700 AI agents created by OpenAI carried out the attack and in many cases tried to cover their tracks. And last week, it was revealed OpenAI agents had also hijacked a German website https://www.reuters.com/world/europe/openai-agents-hijacked-german-website-previously-undisclosed-ai-breakout-this-2026-09-04/ this spring and turned it into a message board for AI agents. Anthropic, meanwhile, has disclosed four instances https://www.cbsnews.com/news/anthropic-ai-model-internet-hack-fourth-time/ of its Claude models hacking external systems. The RubyGems revelation comes at the end of a week of intense scrutiny on AI platforms and calls to pause development until stricter safety standards can be put in place. On Tuesday, an Anthropic researcher announced his resignation from the company on social media, warning https://www.theguardian.com/technology/2026/sep/09/anthropic-researchers-ai-human-extinction that AI could kill off humanity within the next decade. The warnings, echoed by other Anthropic researchers, sparked calls https://www.theguardian.com/technology/2026/sep/09/lawmakers-blast-ai-human-extinct-2030 across the political spectrum for immediate action on AI.