In August last year, the Vanderbilt Institute of National Security published 400 pages of internal documents that were leaked from the Chinese company Golaxy. The documents revealed a sophisticated state-linked operation which uses AI agents to automate the targeting, production, and distribution of propaganda overseas. They showed how Golaxy is using extensive datasets to produce targeted propaganda — including 23 million Taiwanese household registration records, all sitting US Congress members and more than 4,000 other US political and public figures.
This was not an isolated incident: It is a warning about how the economics of foreign influence campaigns are changing.
Advances in AI are not only making these campaigns cheaper to run, but are enabling new tactics that democracies must learn to fight. They lower the barriers to entry, allowing more actors to create and distribute disinformation. They enable communities to be profiled and targeted with a level of precision that was not possible before, reducing trust and increasing divisions. They also change how often these campaigns can be carried out. Until now, such attacks have required human input and peak around politically significant events such as elections. Autonomous agents remove those constraints and theoretically allow them to run continuously over long periods of time.
Many democracies see Taiwan as a model for responding to these threats. Taiwan’s response is largely led by a network of civil society organizations — such as the Taiwan Information Environment Research Center, Doublethink Lab and Taiwan FactCheck Center — which monitor trends, correct false statements and educate citizens. This ecosystem is fast and flexible: it was central to countering Chinese influence operations during the 2024 presidential election. However, it is working under pressure. Last year’s global funding cuts have already forced regional groups such as the Asia Fact Check Lab to close. Taiwan is skilled at identifying and responding to individual attacks, but that alone is not enough. One practitioner has warned that relying on a reactive strategy could limit Taiwan to “doing autopsies” after the damage has already been done.
As the 2028 presidential election approaches, these threats would only multiply. This raises difficult questions about the sustainability of Taiwan’s response.
There are two main changes Taiwan could make to better prepare for operations like Golaxy’s. The first is to expand its detection and correction capabilities. This is what Doublethink Lab did in the 2024 election, hiring short-term fact-checkers to manage a surge of disinformation in social media. In the future, AI tools and partnerships with the private sector might help to grow these capabilities to some degree. However, it is harder to rapidly scale operations across a decentralized system. As the volume and complexity of these threats increase, it might not be feasible to try to counter them one by one.
Second, Taiwan could invest in measures that make people and systems less vulnerable to such operations. This means not just improving media literacy education, but regulating platforms to require transparency and accountability and commissioning research to identify specific areas of vulnerability. Taiwanese practitioners have been calling for greater investment in these types of preventative measures for some time. Groups such as Taiwan Factcheck Center, the Pang-phuann Association of Education and community universities are already delivering targeted media literacy training — but these programs are limited in scale and leave major gaps in coverage.
Realistically, Taiwan needs to pursue both approaches to keep pace. However, neither option is possible without significant government support. Agencies such as the Ministry of Digital Affairs provide some backing for Taiwan’s information resilience efforts, but their role in countering foreign influence is still limited. Taiwan’s civic organizations have long relied on international donors rather than state funding. These arrangements have protected the independence of those organizations and partially shielded them from intense cross-party disputes over Chinese interference in Taiwan. However, they have severely constrained the system’s ability to grow in the long term.
There might be other ways for Taiwan to adapt to these threats. If the government took the lead in building citizens’ social and cognitive resilience, this would have wider benefits for Taiwan’s social trust and cohesion and could be framed in ways that could cut across party divides.
It could consider the approach taken by France, which has created a dedicated agency to counter foreign interference and developed a national strategy to build people’s resilience to such threats. In Taiwan, such a move would require slow and patient coalition-building. It would also require considerable political will to step away from the status quo.
If the government does not take action, then the onus would continue to fall on citizens and communities to do more to protect themselves.
Foreign propaganda operations are increasingly adopting subtler persuasion methods and are grounded in popular culture rather than explicitly political content. This means that alongside crowd-sourced fact-checking tools — such as CoFacts and Auntie Meiyu — informed citizens’ networks and civic advocacy platforms such as the Plainlaw Movement would become even more important. However, these grassroots models have limits — and continuing to rely on them alone might not be sustainable.
Tom Wilson is a public governance practitioner based in Taipei whose work focuses on democratic resilience in the Asia-Pacific. Previously, as a specialist advisor at New Zealand’s Public Service Commission, he led the design of new public agencies and developed mechanisms for cross-government coordination and partnerships with NGOs and Maori organizations.
