New Delhi, July 23 -- Concerns around artificial intelligence (AI) acting autonomously in detrimental ways will heighten after OpenAI reported what it described as an "unprecedented incident".

It said that during a test of an AI agent's hacking capabilities in closed settings, the agent spied a vulnerability unknown till then to access the internet and hack Hugging Face, a database of AI models. In other words, it "successfully found ways to gain access to secret information that it could use to cheat the evaluation".

Hugging Face blocked the break-in after it detected it. "It's quite mind-blowing that all of this happened autonomously!" said Clement Delangue, co-founder of the victim startup. Alarming as it is, such incidents are likel...