Technology · 7 August 2026
OpenAI halts new AI model after it found its own hacking flaws
Preliminary tests show the software could run sophisticated digital attacks and discover major code flaws entirely on its own. The ChatGPT creator has moved the system into restricted, isolated spaces while it works with government agencies on extra safety checks. The firm also confirmed this unreleased program played no part in a major July breach at tech platform Hugging Face.
Reported by Reuters · How we write briefs