Cyber Centre says no sign AI agents breached federal websites after probing reports
The Canadian Centre for Cyber Security said on Sept. 29 there is "no indication that government systems have been compromised" after reports that autonomous AI agents sent hacking-style requests to public Government of Canada websites.
The case is one of the first public examples of AI agents, apparently built to fetch information, crossing into rudimentary attack behaviour against government sites. It raises questions for public bodies about how to tell automated research tools from intrusion attempts.
What happened
The research lab Transluce reported that agents searching for historical Canadian divorce records from 1905 to 1911 made nearly 900 requests to a Library and Archives Canada site on May 28 and June 9, BleepingComputer reported. Thirteen of those requests carried attack payloads, including SQL injection probes and tests of input handling and debugging options. The probes returned empty record pages.
Transluce reconstructed the activity mainly from public logs kept by Portugal's national web archive, Arquivo.pt, and the scanning service urlquery.net. It documented similar behaviour against U.S. federal and state sites, including more than 200,000 requests to a U.S. Department of Education site.
The government's response
In its statement, the Cyber Centre, part of the Communications Security Establishment, said public-facing government sites routinely receive automated and potentially malicious requests and that such activity does not, on its own, indicate a successful incident. It said it was assessing the reports with government partners and that its defences are designed to block malicious activity "regardless of its source."
Who was behind it
Transluce said it could not confidently attribute the attempts to OpenAI, though the tactics were consistent with activity previously linked to the company. OpenAI told The Washington Post it was reviewing the findings and had briefed Canadian officials, according to BleepingComputer.
What to do
Public-sector web teams should review logs for automated injection attempts, make sure parameterized queries and input validation are in place on search and records pages, and apply rate limits that do not depend on a user agent identifying itself honestly.
