Taiwan bans the use of DeepSeek in government agencies, citing security concerns

TAIWAN: Taiwan has prohibited its public sector workers and those employed in key infrastructure facilities from using DeepSeek, citing concerns that it is a Chinese product that could endanger national security.

DeepSeek launched its R1 chatbot last month. It boasts capabilities comparable to leading AI models in the United States yet at a significantly lower investment cost. This has raised international concerns, with several countries, including South Korea, Ireland, France, Australia, and Italy, questioning the Chinese AI startup’s data management practices.

Additionally, it has failed multiple researchers’ safety tests, including those targeting cybercrime, misinformation, illegal activities, and general harm.

Cisco, the team that conducted these tests, expressed their concern. “The results were alarming: DeepSeek R1 exhibited a 100% attack success rate, meaning it failed to block a single harmful prompt,” Cisco stated in an article published by PCMag. “This contrasts starkly with other leading models, demonstrating at least partial resistance.”

Taiwan’s Ministry of Digital Affairs issued a statement

On Friday, Taiwan’s Ministry of Digital Affairs urged government agencies and critical infrastructure sectors to refrain from using DeepSeek. The ministry cited concerns over national information security, specifically regarding the potential for cross-border data transmission and the risk of information leaks, as reported by The Manila Times.

Taiwan has long accused China of employing tactics that are not overtly aggressive but still undermine Taiwan’s security and sovereignty. Since 2019, Taiwan has restricted government agencies from using ICT products and services that might threaten national information security.

DeepSeek faces international scrutiny

Its chatbot has caused a stir on Wall Street, with reports suggesting that its powerful AI capabilities rival US companies at a fraction of the cost. This development has raised further questions, especially given the US’s strict sanctions preventing Chinese firms from accessing the advanced chips required to power such AI models.

Following Taiwan’s move, data protection regulators in South Korea and Ireland have indicated they will seek clarification from DeepSeek on how it handles user data. Earlier this week, Italy initiated an investigation into the R1 model, blocking it from processing data belonging to Italian users.

Read related: US investigates if China’s ‘DeepSeek’ sourced Nvidia chips through Singapore intermediaries