Australia’s prime minister said an artificial intelligence agent developed by ChatGPT-maker OpenAI hacked into a government health care website in June, the latest example of a cybersecurity breach perpetrated with cutting-edge AI.
There was no evidence that personal health information was stolen, but an investigation is ongoing, Prime Minister Anthony Albanese said at a Wednesday news conference while attending the United Nations General Assembly in New York.
The website targeted was an online portal for the country’s Medicare public health insurance system.
The incident adds to a growing list of known cybersecurity breaches involving AI. That this one targeted a government agency in a key U.S. ally is likely to add fuel to ongoing debates over whether AI companies need to be regulated more strictly or if the industry needs to slow down its pace of development.
Albanese criticized OpenAI for the incident and said it took too long for the company to disclose it to the Australian government. He said he personally spoke to OpenAI chief executive Sam Altman about the hack on Wednesday.
“This situation is obviously unacceptable,” Albanese said. “Today, I spoke with the CEO of OpenAI, Sam Altman, to express Australia’s extreme concern about this incident.”
OpenAI spokesperson Drew Pusateri said in an emailed statement that the company’s AI models “took actions we did not intend” while looking up information on Australia during an internal test of their capabilities. “Our review found no evidence of patient records being accessed,” the statement said.
Altman appeared at the U.N. on Wednesday alongside other AI company leaders, who called for international cooperation on regulating the technology.
In July, OpenAI disclosed that some of its AI agents had escaped the company’s systems, accessed the internet and hacked into another AI company’s servers. The disclosure was followed by admissions from other companies, including Anthropic, Meta and Google, that some of their own agents had hacked into other companies too.
Independent researchers have since found other incidents of OpenAI’s agents surfing the web independently.
At the same time, criminals and government hackers have begun employing AI tools to supercharge their own efforts. AI has allowed hackers to move much more quickly, and cybersecurity firms have been scrambling for months to shore up defenses and try to contain the new capabilities.
The post Australian PM says ChatGPT-maker’s AI hacked public health care site appeared first on Washington Post.
