Federal government bans Chinese AI startup DeepSeek on public service devices

Article content

OTTAWA — The federal government is barring the use of Chinese artificial intelligence startup DeepSeek’s chatbot on government devices over “serious privacy concerns” the company can inappropriately collect and retain sensitive data.

In a memo to government heads on Thursday, the government’s Chief Information Officer Dominic Rochon said the Chinese-owned app — which unexpectedly became one of the world’s most downloaded last week — is now banned on devices and networks overseen by Shared Services Canada (SSC).

Article content

He also called on heads of departments and agencies that don’t use SSC devices and networks to block the application and website on their networks.

“Due to serious privacy concerns associated with the inappropriate collection and retention of sensitive personal information, and as a precautionary measure to protect government networks and data, it is recommended that departments and agencies restrict the use of the DeepSeek chatbot on government devices,” Rochon wrote in the memo obtained by the National Post.

The ban does not extend to Canadians’ personal devices.

DeepSeek shocked the tech world last month when the small Chinese startup launched its AI model that reportedly matched or surpassed existing products, such as OpenAI’s ChatGPT, in various benchmarks, including programming and reasoning.

Its emergence proceeded to disrupt the stock market, causing significant drops in the value of AI-related companies like Nvidia. In a short period, the Chinese company’s success has reignited the AI race between China and the United States, challenging U.S. leadership in the sector.

Article content

Article content

But it was also controversial for how it applied Chinese censorship laws to some of its responses. For example, the app ignores questions about the 1989 Tiananmen Square massacre.

Even more concerning for Western governments is how the company handles users’ personal information in light of sweeping Chinese national security laws that allow its government to compel companies in China to hand over data.

Those concerns are what led the Canadian government to ban TikTok, owned by Chinese tech giant ByteDance, on public service networks and move last fall to force TikTok Canada to dissolve.

In recent days, South Korea, Australia and Italy have blocked DeepSeek’s chatbot from government devices and many other countries are mulling similar moves, including the U.S.

In his Thursday memo, Rochon also said his office is reviewing its policies on government device management to ensure they are up to date. He also said SSC is developing a “more rigorous approach” to how it manages government devices under its purview.

The federal government did not immediately respond to questions about what threats it detected from DeepSeek.

Article content

Former national security analyst Dennis Molinaro told National Post that if an app is a potential threat to a government device, then it’s not safe for personal use either.

He also warned that using the app is risking handing over your data to the Chinese Communist Party’s (CCP) regime for AI training.

“I would argue this is far worse than Tiktok because AI researchers know how valuable data is to training devices, so the hoovering up of data to be sent back to China is more significant and more important,” said Molinaro, now a legal studies professor at Ontario Tech University in Oshawa.

“The west is in a race when it comes to AI development. By using this app, you are assisting the CCP with data collection and training its AI which could be used one day against the west.”

Last week, AI security researchers at Cisco found that DeepSeek’s AI chatbot failed all of the testers’ cybersecurity tests.

“Our findings suggest that DeepSeek’s claimed cost-efficient training methods… may have compromised its safety mechanisms. Compared to other frontier models, DeepSeek R1 lacks robust guardrails, making it highly susceptible to algorithmic jailbreaking and potential misuse,” they wrote.

Article content

National Post, with additional reporting from Stewart Lewis
cnardi@postmedia.com

Get more deep-dive National Post political coverage and analysis in your inbox with the Political Hack newsletter, where Ottawa bureau chief Stuart Thomson and political analyst Tasha Kheiriddin get at what’s really going on behind the scenes on Parliament Hill every Wednesday and Friday, exclusively for subscribers. Sign up here.

Our website is the place for the latest breaking news, exclusive scoops, longreads and provocative commentary. Please bookmark nationalpost.com and sign up for our daily newsletter, Posted, here.

Article content