OR Microsoft He announced that he had identified hackers based in US and abroad, who bypassed the protective rails in artificial intelligence tools – including the company’s Azure Openai services – to create harmful content, including non -consensual images of celebrities.

According to the company, hackers removed customer connections from public sources and used them to access artificial intelligence productive services, including Azure OpenaiMicrosoft’s Cloud product that allows customers to use OpenAi models. The hackers then changed the capabilities of artificial intelligence products and sold other malicious groups, giving them instructions on how to create harmful content.

The hackers found by Microsoft, outside the US, are based in Iran, the United Kingdom, Hong Kong and Vietnam. They are alleged to be part of a global cybercrime network that Microsoft calls Storm-2139. Two other members are in Florida and Illinois, but Microsoft said he is not named them to avoid derailment of criminal investigations. The software manufacturer said he was preparing criminal referrals to US and foreign law enforcement authorities.

“We take very seriously the abuse of artificial intelligence, recognizing the serious and permanent impact of abusive images for victims,” ​​said Thursday. RugAssistant General Advisor to Microsoft Digital Crimes Unit, in its post.

“Microsoft remains committed to protecting users, incorporating strong AI security measures on our platforms and protecting our services from illegal and harmful content,” he said.