Thomas Wolf (토마스 울프), a co-founder of open-source AI model-sharing platform Hugging Face, called an incident in which OpenAI’s top AI model went out of control and hacked Hugging Face a warning to the industry, the BBC’s website reported on Tuesday.

Wolf said on the BBC radio programme “Newsday” that such cyberattacks would become the most common in the future. He added that most companies still do not know that the rules of the game have changed, the report said.

OpenAI said on Monday that its AI model left a test environment and carried out a cyberattack. It called the incident unprecedented and said it was investigating with Hugging Face.

Wolf said that when signs of the attack first appeared in mid-July, he was unable to identify the source at all. He said he ultimately prevented the damage from spreading.

Wolf said the breach was completely different from attacks he normally experienced. He added that OpenAI quickly informed Hugging Face that its model was behind it. Wolf, who also serves as Hugging Face’s chief scientific officer, added that in a very short time, 17,000 attacks from multiple IP addresses flooded into the Hugging Face network.

Nate Soares (네이트 소어리스) of the Machine Intelligence Research Institute said the hack was concerning and that OpenAI’s model appeared to have ignored safeguards designed to prevent cyberattacks. He added that the model would have known to some extent that this differed from the developers’ intentions, but did not care.

A British government spokesperson said the country’s AI Security Institute was investigating how the AI system operated in the incident. The spokesperson said the government was continuing work to strengthen safeguards together with several laboratories including OpenAI.