San Francisco: (RightNow) A Meta artificial intelligence (AI) model hacked a company’s system during cybersecurity testing. The incident occurred when a partner’s mistake allowed the AI model internet access.

Meta clarified in a statement that the error was made by an independent testing company, Irregular. Investigations into the incident are ongoing.

The incident came to light after OpenAI and Anthropic also admitted their AI models had hacked company systems during testing.

This incident highlights security risks of AI models and poses challenges for developers. Experts suggest such events emphasize the need for caution in technology advancement.

It is noteworthy that OpenAI had previously accessed Hugging Face company accounts through their models. Anthropic has also confirmed similar incidents.