
On June 26, 2025, Kai-Shen Huang, Director of the Democratic Governance Program at DSET, and Non-resident Fellow Yi-Xiang Sun delivered keynote presentations at a roundtable held at the European Parliament. Using Taiwan as a case study, they outlined new trends in Foreign Information Manipulation and Interference (FIMI), including its evolution into tangible real-world threats and the emergence of AI as a new domain for FIMI. They also explored how democratic countries worldwide can enhance collaboration in developing AI talent.
This event was organized by Saman Nazari, Lead Researcher at Alliance4Europe, in partnership with Members of the European Parliament (MEPs) from the Inter-Parliamentary Alliance on China and MEP Markéta Gregorová. The roundtable, hosted within the European Parliament, featured insights from DSET’s latest research on democratic governance.
Huang highlighted how foreign social media platforms have increasingly penetrated the daily lives of many young people. He noted that the algorithms behind these platforms are often influenced by authoritarian values, using entertainment to attract audiences and subtly reinforce FIMI narratives. Citing research by Taiwan’s Doublethink Lab and IORG, Huang pointed out that some Taiwanese users of foreign platforms are more susceptible to authoritarian narratives. Although Taiwan’s Executive Yuan has banned these platforms on official devices, public use remains unrestricted, allowing foreign platforms to continue shaping Taiwan’s information environment.
Huang further emphasized that FIMI is no longer confined to the virtual realm but is manifesting as real-world threats. He cited the Southport riots in the UK as an example of how disinformation can rapidly spark social unrest, and referenced Taiwan’s previous “Indian migrant workers” rally to illustrate a similar emerging trend domestically.
The rapid advancement of AI technologies has also opened new avenues for FIMI. Huang explained that authoritarian states may exploit AI in two ways: one is through “model export,” training AI with censored data so that the models inherently carry embedded censorship; the other is through “data infiltration,” whereby inappropriate data is inserted into training datasets, effectively “infecting” AI models in democratic countries. Both approaches erode AI with authoritarian values, turning it into a crucial node for FIMI activities.
In response, Huang stressed that Taiwan and other democratic nations must reassess their strategies for countering authoritarian AI. The current AI race has expanded beyond just building models to developing practical solutions. Small and medium-sized democracies should avoid becoming overly fixated on technological nationalism and instead promote AI policies that align with their unique strengths. Furthermore, beyond export controls, more attention should be paid to the potential risks posed by inbound AI solutions from abroad.
During the event, DSET Non-resident Fellow Yi-Xiang Sun also shared his recent research on AI talent policies in the US, EU, and Taiwan. He noted that the US primarily attracts global high-tech talent through programs like H-1B visas and the CHIPS Act, while the EU focuses on transnational academic and industry collaboration, strengthening regional AI talent networks through initiatives such as ELISE and EIT AI. In contrast, Taiwan’s efforts have emphasized local talent cultivation and technical practice, but there is still room for improvement in international mobility and cooperation.
As reflected in the event’s theme — “Cross-Border Channels, Cross-Democracy Risks” — the FIMI challenges facing Taiwan’s democratic resilience are not unique but shared by democracies worldwide. Through continued sharing and dialogue, Taiwan can contribute valuable experiences and insights, building partnerships with other democracies to collectively address the new threats that AI and FIMI pose to democratic resilience.