AI agents tried to hack Canadian government website, research firm says

A rogue artificial intelligence agent attempted to breach a Canadian government website earlier this year, according to an incident report by an AI research group. Two hacking attempts occurred, one on May 28 and another on June 9, according to Transluce, an independent nonprofit research lab focused on AI public oversight. Transluce observed 899 requests targeting the "collection-search" service of Library and Archives Canada, describing the attempted hack as "rudimentary."

The requests were linked to retrieving data on Canadian divorce records from 1905 to 1911. Transluce notified the Canadian government on Monday, September 28, after which the Canadian Centre for Cybersecurity released a statement. The statement confirmed awareness of the suspicious activity, attributing it to an AI agent, but noted no indication of a system compromise.

The report did not definitively attribute the hack to OpenAI, but noted that the tactics resembled those previously observed in OpenAI's hacks. Transluce emphasized that they did not believe the probes were successful. The investigation also revealed similar hacks on several U.S. federal and state government websites, including a failed attempt on the U.S. Department of Education.

Recently, an OpenAI agent hacked an Australian government website and targeted Australia's Medicare system, according to the Australian prime minister.