A road crossing in the Xinyi district of Taipei, Taiwan in January 2025. ABS-CBN News, file

TAIPEI — Taiwan has banned workers in the public sector and at key infrastructure facilities from using DeepSeek, saying it was a Chinese product and could endanger national security.

DeepSeek launched its R1 chatbot last month, claiming it matches the capacity of artificial intelligence pace-setters in the United States for a fraction of the investment.

Countries including South Korea, Ireland, France, Australia and Italy have raised questions about the Chinese AI startup's data practices.

EXPLAINER: Could DeepSeek spark an AI race between US and China?

Taiwan's Ministry of Digital Affairs said Friday all government agencies and critical infrastructure should not use DeepSeek because it "endangers national information security".

"DeepSeek AI service is a Chinese product," the ministry said in a statement.

"Its operation involves cross-border transmission and information leakage and other information security concerns."

Taiwan has long accused China of using "grey zone" tactics — actions that fall short of an act of war — against the island, including cyberattacks, as Beijing presses its claims of sovereignty over the island.

Since 2019, Taiwan has banned government agencies from using information and communication technology products and services that pose a threat to "national information security".

DTI to boost digitalization, AI technology use for consumer protection

DeepSeek sparked panic on Wall Street this week with its powerful new chatbot that is thought to have matched US companies in its abilities but at a fraction of the cost.

That's despite a strict US regime prohibiting Chinese firms from accessing the kinds of advanced chips needed to power the massive learning models used to develop AI.

Taiwan's restriction came as data watchdogs in South Korea and Ireland said they would ask DeepSeek to clarify how it manages users' personal information.

Earlier this week, Italy launched an investigation into the R1 model and blocked it from processing Italian users' data.

© Agence France-Presse