Taiwan warns on DeepSeek, other Chinese AI apps over security, bias

Taiwan’s National Security Bureau has warned users about security and bias risks linked to five Chinese generative AI apps after a recent inspection.

The bureau, together with the Ministry of Justice Investigation Bureau and the Criminal Investigation Bureau, reviewed Deepseek, Doubao, Yiyan, Tongyi, and Yuanbao for issues across categories including data collection, permissions, and biometric access.

All five apps were found to violate multiple security indicators, with Tongyi failing 11 out of 15.

Officials said the apps request location data, collect screenshots, and require acceptance of privacy terms that may compromise user data.

The content generated by these apps was also found to reflect Chinese government positions on cross-strait issues and avoid politically sensitive terms.

Taiwan has banned Deepseek from use on government devices since February, but no ban exists for the other apps or private use of Deepseek.

🔗 Source: Focus Taiwan

🧠 Food for thought

Implications, context, and why it matters.

Taiwan’s opaque audit methods leave gaps

Auditors cited violations on 15 security indicators. Agencies have not disclosed the indicators, the tests, or which data flows raised alarms. Without that detail, Taiwan-based companies cannot judge if these apps carry more risk than popular foreign apps that also ask for location data and screenshots.
Content aligned with Chinese government positions and avoided politically sensitive terms in the findings. Taiwan is rolling out Large Language Models (LLMs) and multimodal AI. These systems process text, images, and other inputs. Uses include regulatory compliance plus medical education. The risk picture spans bias, influence efforts, plus technical flaws. The report does not spell out that split.

Local Mobile Device Management (MDM) and Data Loss Prevention (DLP) vendors can meet near-term demand

Taiwan’s Internet of Things (IoT) sector topped New Taiwan dollars (NT$) 1 trillion with Information and Communications Technology (ICT) services reaching over 80%. That base creates many companies with corporate devices and Bring Your Own Device (BYOD) policies that need app risk controls. These buyers want MDM and DLP tools that detect and block flagged Chinese AI apps without waiting for a blanket ban.
COVID-19 pushed remote work, which widened openings for data leaks through employee-installed apps. Cybersecurity suppliers in Taiwan that ship app intelligence capabilities fast can stand out. Those features include finding which Chinese AI apps staff use and limiting access to corporate data, with rules tuned to Taiwan’s regulations.