AI agents are back in the headlines for all the wrong reasons and this time the target was Canadian government websites. According to research firm Transluce, AI agents attempted to breach Library and Archives Canada on two occasions this year. Canada’s federal cyber agency has issued a statement saying it found no evidence that its systems were compromised.
What happened with the Canadian government site?
Transluce found the agents tried to access the Canadian site on May 28 and June 9, sending hundreds of requests that included a series of apparently failed, rudimentary hacking attempts. It notified Canada’s government on Monday, and the Canadian Centre for Cyber Security confirmed Tuesday it was looking into the reports.
Transluce says the tactics resemble prior activity it has linked to OpenAI, but it isn’t confident enough to pin this specific incident on the company. According to Reuters, OpenAI has only acknowledged being aware that its models attempted to reach publicly available information on Canadian government sites.
A pattern that keeps on repeating
The Canadian incident arrives shortly after an OpenAI agent breached an Australian government health data portal and accessed files. OpenAI agents also reportedly probed US government systems without successfully breaking in. Then there is the Hugging Face incident, where OpenAI’s agents escaped a controlled testing environment and carried out a large-scale intrusion.
OpenAI said the US activity mostly involved routine tasks, like accessing public web content, that drifted past a line nobody drew clearly enough. The company is still investigating the full scope of this behavior and has paused training on advanced new models while it does.
For now, Canada appears to have avoided a breach, but the repeated attempts suggest government networks around the world need to brace for the growing threat posed by increasingly autonomous AI agents.
RELATED COVERAGE:
- OpenAI has a rogue AI problem and ChatGPT users are now caught in it
- OpenAI AI agents bombarded the United Nations with aggressive tactics to extract data
- Nvidia has a fix for rogue AI agents to prevent incidents like the Hugging Face hack
- OpenAI’s rogue AI agents scoped out Hugging Face months before the big breach
- OpenAI is investigating more incidents of AI agents going rogue days after hack
- Experts used AI to break into OpenAI. They say it was cheap and a red alarm for the industry
