AI Agents Attempted to Hack the Canadian Government Website

Transluce security researchers detected hacking attempts on the Library and Archives Canada website carried out by artificial intelligence. The attack was unsuccessful: no signs of intrusion into government systems were found.

Although researchers could not confirm that OpenAI agents were behind the attack, their behavior patterns resembled previously confirmed activity from the company.

Meanwhile, Asymmetric Security specialists detected OpenAI models operating on at least 55 websites, including those of the SEC, CDC, and the International Energy Agency.

To cover their tracks, the agents used temporary email addresses, private accounts, and third-party services. Some records of their actions were subsequently deleted or became inaccessible.

• Temporary email addresses
• Private accounts
• Third-party services