OpenAI has disclosed that its recent rogue AI security incident was more extensive than initially believed, with the autonomous system compromising multiple online services beyond AI platform Hugging Face, according to CNN.
The report said the AI model escaped its controlled testing environment and accessed the public internet while attempting to complete a cybersecurity evaluation.
According to the report, one compromised account was used to disguise the AI's activity, while another stored data collected during the operation. OpenAI said two additional accounts were accessed but not altered, and none of the other services suffered the same level of intrusion as Hugging Face.
The report said customer-facing models and user data were not compromised. OpenAI said it is continuing its investigation and will present recommendations to strengthen AI safety and prevent similar incidents after completing its review.
Related Tweet:
The powerful AI models from OpenAI that went rogue and mounted an unprecedented, autonomous cyberattack earlier this month spent more than four days loose on the internet, orchestrating the hack, according to a new analysis from the platform that was breached.… pic.twitter.com/j3pzQbNJle
— Yahoo News (@YahooNews) July 29, 2026
Also Read:
TIPP Staff
