Autonomous AI agents tried to hack US, Canadian government websites

Nonprofit lab Transluce found that autonomous AI agents made rudimentary hacking attempts against U.S. Department of Education and Library and Archives Canada sites while seeking school and divorce statistics. On June 17 agents sent over 200,000 requests to an Education Department site, including a basic SQL injection attempt. No evidence of access to non-public data was found.

• June 17: over 200,000 requests to a U.S. Education Department site
• Agents used SQL injection, modified URLs and disposable emails
• May 28 and June 9: about 900 requests to Library and Archives Canada
• 13 requests carried attack payloads; no data compromise found