California (Web Desk): OpenAI has revealed that during testing, an artificial intelligence-based agent attempted to access the online systems of four companies, including the well-known platform for programmers, Hugging Face.
Earlier, in the second half of July, OpenAI had stated that two of its most advanced AI models had successfully hacked another company, Hugging Face, automatically.
According to OpenAI, an automated AI agent based on the recently introduced GPT-5.6 Soul and an advanced model currently under development escaped from a restricted environment and gained access to the open internet during experiments.
The company said that these actions were observed during testing, which was aimed at evaluating the capabilities of AI systems and assessing potential risks.