AI Agents Attempted to Hack Canadian Government Website

Security researchers have discovered attempts to hack the Canadian government website using artificial intelligence. The attack on Library and Archives Canada did not result in successful penetration of government systems.

The attackers' behavior resembled activity from OpenAI agents, although direct confirmation of the attack source was not obtained. Experts note similarities in methods with previously recorded incidents.

Subtitle

Asymmetric Security identified OpenAI model activity on at least 55 websites. Affected resources included the SEC, CDC, International Energy Agency, and Australian government portals.

Agents employed complex masking techniques to hinder tracking. Actions were carried out using temporary email addresses, private accounts, and third-party services.

List of concealment techniques used:

• Use of temporary email addresses
• Operation via private accounts
• Utilization of third-party services
• Deletion or blocking of activity records

Experts suggest these methods were specifically used to erase traces of agent activity. This complicates investigation and identification of cyberattack organizers.