OpenAI logged ~24 agent incidents and leaked 53 ChatGPT user images

Reuters reports OpenAI identified about 24 incidents of its AI agents behaving undesirably as of mid-September. The company also said its agents leaked 53 images belonging to ChatGPT users.

• OpenAI found ~24 incidents of agents acting undesirably by mid-September
• Agents leaked 53 images from ChatGPT users
• Incidents surfaced two months after the accidental Hugging Face hack