Following a report stating Canadian government systems were targeted by AI agents, Canada’s digital security agency says there’s no indication anything was compromised.
The news: Earlier this week, AI research firm Transluce said it discovered several incidents where AI agents appeared to use “aggressive techniques” to access publicly available data on Canadian government websites. This included an attempted attack against Library and Archives Canada (LAC), which the firm said failed.
Transluce said it alerted the Canadian Centre for Cyber Security, which confirmed in a public statement that there is no indication that government systems were compromised “at this time.”
From the source: Transluce’s report said it captured 899 requests hitting the “collection-search” service of LAC associated with retrieving data on divorce records in Canada from between the years 1905 and 1911. However, 13 of these requests carried “attack payloads” that were looking for vulnerabilities.
Communications Security Establishment Canada said that public-facing government websites routinely receive requests like this, and that it’s an “ongoing feature of the online environment and does not, on its own, indicate a successful cyber incident.”
The context: Following a high-profile agentic-AI-powered cyberattack involving OpenAI and Hugging Face over the summer, multiple AI firms, including Google, Anthropic, and Meta, have said that AI agents powered by their models have performed cyberattacks. This includes government systems, as Australia disclosed last week that an OpenAI agent hacked into its national healthcare database.
Transluce said it does not “confidently” attribute the LAC breach attempt to OpenAI, but that it did exhibit tactics similar to OpenAI’s agents around a similar timeframe as other incidents.
Final thought: The attempted breach comes as international cybersecurity players, including the Canadian Centre for Cyber Security, are urging organizations to boost cyber defences, as rapidly advancing AI models are proving to be adept at finding vulnerabilities. According to the 2026 Microsoft Digital Defense Report, released on Thursday, AI is significantly speeding up cyberattacks, and the majority of cases involve some kind of data theft.
With files from Madison McLachlan.
Feature image courtesy Wikimedia Creative Commons. Shared under license CC0 1.0.