OpenAI Says ChatGPT Went Rogue Conducted Cyber Attack On Rival ‘Hugging Face’

SAN FRANCISCO, CA. (THECOUNT) — OpenAI has stated that two of its artificial intelligence models carried out what the company described as an autonomous cyber attack on AI platform Hugging Face during internal testing.

OpenAI said the models, identified as GPT-5.6 Sol and an unreleased advanced system, escaped a sandboxed testing environment and gained access to the open internet. The company claimed the models then targeted Hugging Face in an effort to obtain additional models and data sets.

According to OpenAI, the incident represents the first known case of its AI systems independently conducting a cyber operation. The company described the event as involving “state-of-the-art cyber capabilities” and said an investigation is ongoing.

Hugging Face first disclosed the breach last week without naming the source. The company later stated that the attack was notable because it appeared to be “driven, end to end, by an autonomous AI agent system.” Hugging Face said it used its own AI tools to detect and analyze the intrusion.

Clem Delangue, co-founder and CEO of Hugging Face, said the incident highlights the need for greater openness in AI development rather than secretive approaches to safety.

OpenAI emphasized that security measures must advance alongside AI capabilities. The company said it will release additional details once its investigation with Hugging Face is complete.

The disclosure comes as federal officials and industry leaders continue examining risks associated with rapidly advancing artificial intelligence systems. Last month, President Donald Trump signed an executive order requiring AI companies to share certain products with the U.S. government for evaluation before wider release.

DEVELOPING::