Join War on the Rocks and gain access to content trusted by policymakers, military leaders, and strategic thinkers worldwide.
Taiwan produces the world’s most advanced semiconductors, yet the island’s own military is only in the opening stages of adopting artificial intelligence systems. This mismatch bears directly on Taiwan’s ability to deter Chinese aggression.
Beijing is pushing to develop and field sophisticated AI-enabled military capabilities. Given China’s focus on “reuniting” with Taiwan, the Taiwanese military could be the first to contend with the People’s Liberation Army’s emerging defense technologies during a potential conflict. Should the United States, and possibly allied forces, come to the island’s defense, a coalition would benefit from having access to AI-enabled military systems that share data and operate together in a conflict. Therefore, Taiwan’s military AI readiness is critical to its own national security and could enable the execution of more effective operations with partner countries.
Taipei appears to recognize the importance of military AI innovation. Taiwan’s Ministry of National Defense has signaled its plans to procure emerging capabilities. While the ministry has historically not followed through on many of its stated reforms, it is now backing up its words with actions. Since 2024, it has established the Defense Innovation Office and a dedicated AI office. Though in its early stages, the latter is focused on accelerating the acquisition of AI applications, such as those related to image recognition and cyber defense. Recently, the Ministry of National Defense began encouraging individual services to adopt AI tools. It has also stated its intention to establish an “AI Governance Committee” to promote AI risk classification, development verification, and security monitoring frameworks. These are welcome developments, but the extent to which Taiwan’s military is prepared to develop and adopt emerging AI capabilities remains unclear.
Ultimately, the Ministry of National Defense should refocus its AI office on force-wide governance and clear technology priorities rather than adoption alone, tying together data, industry, and talent. Realizing this vision will require the Legislative Yuan to prioritize AI in the annual defense budget and deepen collaboration with partner countries and their defense technology companies. As Taiwan’s two main political parties are notoriously at odds on defense spending, this will require delicate political maneuvering, but it is necessary to bolster the island’s defenses.
We reviewed all of the AI-focused, open-source, Mandarin-language articles published since 2023 by Taiwan’s military journals and the Institute for National Defense and Security Research, the island’s only Ministry of National Defense-linked think tank. Despite the fact that nearly 50 articles on AI have been published in that period, only about one-third of them address the AI-related problems Taiwan is facing. Although Taiwan’s relatively small size might explain this quantity of public commentary, AI still appears to be a relatively under-discussed topic, as the same journals and think tank published nearly 90 articles on unmanned systems during that timeframe. The limited number of articles notwithstanding, we believe they illuminate the issues the Taiwanese military faces as it develops and deploys AI-enabled military systems. Taiwan’s defense community has identified six barriers related to AI governance, digital infrastructure, the data enterprise, cybersecurity, military-civil collaboration, and AI talent.
Taken together, they suggest that while the Ministry of National Defense appears to understand the potential utility of military AI applications, it faces institutional, infrastructural, and technical challenges that will complicate its widespread adoption of military AI capabilities.
AI Governance
The defense experts identify Taiwan’s fragmented military AI governance as a major gap. They argue that the Ministry of National Defense has established neither a unified AI framework nor common technical standards for AI systems, reflecting that Taiwan’s military AI interests are organized around specific use cases rather than coherent technical or operational architectures. Without common guidance, they warn, each branch of the military could develop systems according to its own mission requirements and technical specifications, creating risks of duplicated investments. Skeptics could argue that a more fragmented development pathway might succeed given a long enough timeline. But Taiwan does not have the luxury of elongated timeframes, and stovepiped development, internal competition, and eventual consolidation around a winning governance framework could consume years that the island may not have. Indeed, during a near-term conflict, AI systems developed under siloed governance structures could have limited inter-service interoperability and, therefore, reduced utility. Finally, Taiwan’s governance fragmentation will make it more difficult to integrate with its partners’ AI tools.
Digital Infrastructure
According to several experts, the Taiwanese military’s still-developing digital infrastructure could pose a barrier to AI adoption. They argue that the Ministry of National Defense is struggling to make sufficient investments in basic digital equipment and has established only limited network bandwidth in remote areas, creating difficulties in ensuring reliable communications. Perhaps more significantly, some experts point out that the Ministry of National Defense currently has insufficient cloud storage and computing capacity that could hamstring Taiwan’s military AI capabilities. The ministry should address its cloud deficit before AI systems can be integrated across the Taiwanese military. Otherwise, infrastructure gaps could make these tools difficult to deploy and potentially unreliable.
The Data Enterprise
Similarly, the scholars worry that Taiwan’s military data enterprise is largely fragmented and underdeveloped. The authors emphasize the Taiwanese military’s insufficient data volume, uneven data quality, underdeveloped data management systems, and the siloing of data across units and systems. There is a risk that Taiwan’s current military data challenges will limit the Ministry of National Defense’s ability to collect, organize, clean, and share the data to support widespread AI adoption. This is not a total barrier, but without improved data quality, greater data quantity, and better management practices, Taiwan’s AI tools might be trained on small and poorly structured datasets, making them less useful on future battlefields. Moreover, it is necessary to create data-sharing standards that will require Taipei to redress domestic technical and legal barriers.
Cybersecurity Challenges
The experts also highlight AI-related cybersecurity challenges, reflecting the tension between maintaining cyber defenses via network isolation and ensuring broader AI integration. Several authors point out the risks of information leakage and cyber intrusion created by connecting military systems, data, and AI tools. These concerns align with current Taiwanese protective approaches that emphasize strict hardware controls and air-gapping. These methods, however, are hardly compatible with software-intensive AI systems that require frequent updates and widespread data access, indicating the need to consider tradeoffs between cybersecurity and operational utility. Air-gapping could protect networks given the island is constantly targeted by Chinese intelligence, and Taiwanese military personnel have leaked information to Beijing. But declining to develop a smoother system for pooling data for AI systems could be equally risky. While some Taiwanese experts note the importance of software-based cyber defenses, adopting shared standards will also be necessary. Such a shift, if it ever occurs, will require not only significant cyber policy changes in Taiwan, but also a willingness to experiment with novel cybersecurity approaches.
Civil-Military Collaboration
Next, the experts identify lagging military-civil AI collaboration as a significant challenge to Taiwan’s military modernization ambitions. Recognizing that military AI cannot be developed by the Ministry of National Defense alone, they call for a military-civil collaboration model that combines government guidance, private sector leadership, and the military’s application of emerging technologies. At the same time, some scholars advocate for working closely with leading companies in partner countries as well. Other observers argue that the relationship between Taiwan’s military and private sector is especially weak, partly due to espionage concerns, though the ministry’s growing reliance on a new generation of drone manufacturers suggests the model is beginning to shift. Fielding military AI depends less on making advanced chips, a task at which Taiwan excels, than on innovating defense software, creating robust data pipelines, and integrating models, which are areas where Taipei has underinvested. The recently established Defense Innovation Office and AI Office may improve military-civil integration, but the Ministry of National Defense should continue to develop a coherent strategy to help civilian companies pursue military innovation.
Talent and Training
Finally, the experts note that talent and training gaps represent another major challenge. They point out that Taiwan’s armed forces lack sufficient AI expertise across the Ministry of National Defense’s technical, operational, and organizational levels to support broad AI implementation. While U.S. talent discussions often highlight the need to augment operators’ technical capabilities while also attracting more civilian experts to work with the military, Taiwanese debates focus on better training existing uniformed military personnel. However, many of Taiwan’s soldiers have been busy countering China’s daily coercive activities and conducting heightened training alongside their normal duties. Under these circumstances, there is not yet a clear framework for augmenting operators’ technical skills. Having a limited number of technically proficient military personnel could make AI systems more likely to be superficial technology displays rather than legitimately useful battlefield tools. Furthermore, poor training could lead to the misinterpretation of AI outputs in a contingency. Independent reporting corroborates these concerns, as the Taiwanese military has filled around 80 percent of its total positions. Moreover, international analysts argue that, despite a growing defense budget, redressing underlying personnel and training shortcomings will be difficult, as many of the issues stem from a risk-averse military bureaucracy that has shied away from more rigorous training.
While Taiwan is adjusting to the fast-evolving pace of military AI innovation, the United States is pushing ahead. Despite its early progress, the United States still faces significant challenges, and its defense experts have spilled much ink illuminating them. U.S. discussions of issues related to military AI acquisition cover more ground than Taiwanese debates. For example, conversations among U.S. policymakers and analysts about the tension between conventional hardware-centric and iterative AI acquisition processes have focused on issues ranging from prolonged procurement cycles to inflexible funding restrictions.
These topics do not appear in the Taiwanese articles, but Taipei likely faces similar problems. Taiwan’s current procurement process features manifold bureaucratic hurdles, and the island’s military budgeting process is rife with restrictions that complicate the fielding and updating of software-centric systems. The absence of these discussions, however, is not dispositive evidence, as many of the articles are application-specific, and budget debates may play out outside public journals. Furthermore, the authors seldom offer broader diagnoses of the institutional, infrastructural, and operational barriers to military-wide AI operationalization, as do many U.S. experts. That said, the articles, when analyzed together, suggest that it would behoove Taipei to take a system-wide approach to facilitate a more effective pathway to developing and adopting emerging military capabilities.
These issues are not unique to technology adoption in Taiwan alone. The difficulty of building interoperable AI systems, as evidenced by AUKUS and NATO, lies, at least in part, in the fragmentation of governance frameworks, security or technical standards, computer networks, and data. Therefore, the gaps the Taiwanese experts identify may constitute weaknesses not only in Taipei’s domestic military AI development but also in future coalitional defenses. Early cooperation on these issues could help prevent Taiwan from facing the same challenges that currently constrain other U.S. partners.
Such collaboration would depart from, but likely augment, the conventional model of U.S.-Taiwanese defense ties. Bilateral defense cooperation often focuses on military hardware like drones and tanks, and continued U.S. support for the island is becoming increasingly uncertain. But the United States can play a meaningful role by sharing lessons with Taiwan on AI integration, potentially aiding the military in identifying and addressing barriers before they become more deeply embedded. While some government-to-government exchanges are already taking place, further institutionalized dialogues on both best practices for and mistakes to avoid in developing and deploying military AI responsibly and safely would benefit Taipei. Additionally, should Taiwan develop strong standards around cybersecurity and data sharing, as well as explore legal changes needed to facilitate cross-border military data sharing, it would reduce the barriers to fighting effectively alongside the United States in a future conflict. Of course, Taiwan is not a U.S. treaty ally, and Chinese intelligence penetration of its military remains a persistent problem, both of which complicate the prospects for developing and fielding interoperable systems. Washington also faces its own challenges associated with sharing sensitive AI-related technologies and developing and adopting AI for warfare. But narrowing the gap between the two militaries’ standards could enable more effective operations in a future crisis.
Skeptics might point out that, given unclear U.S. security commitments to Taiwan, developing common standards to facilitate greater interoperability is a fool’s errand. But AI-focused reforms would strengthen Taiwan’s capabilities, so they will pay off whether or not Washington shows up.
The Ministry of National Defense can do more to facilitate military-wide AI integration. First, it should shift the focus of the new AI Office from encouraging AI adoption to issuing guidance that better coordinates adoption across the Taiwanese military. This would include establishing a unified military-focused AI governance framework and clear technology priorities. Such an approach would aid Taiwan in moving from scattered AI experimentation toward a more coherent defense AI ecosystem. Relatedly, fielding AI-enabled systems will be of limited utility if the Taiwanese military does not update its doctrine and concepts of operation to use them effectively. While efforts are underway, critics point out that there is more work to be done here. Moreover, Taiwan’s military AI ambitions will require support from the Legislative Yuan. The defense budget will need to fix issues like the aforementioned digital infrastructure gap. The May 2026 special defense budget, which funded U.S. arms purchases but cut domestic emerging technologies programs, demonstrates why the ministry cannot rely on supplemental funding for these capabilities. Including AI-enabled systems in the annual defense budget process is a better path to maintaining momentum for adoption.
Finally, there is space for greater collaboration between the Ministry of National Defense and U.S. defense tech companies. The recently announced U.S.-Taiwanese cooperation on drone development is a step in the right direction and could point the way toward broader cooperation in emerging technologies.
To be sure, Taiwan’s deeply rooted political, bureaucratic, and organizational issues will likely complicate the execution of these initiatives. Together, however, they would nudge Taiwan toward innovating and adopting emerging capabilities before the next conflict and lay the groundwork for more effective coalition use of emerging technologies.
Fanny Chao is a Taiwan-focused researcher. She is a former student research assistant at Georgetown University’s Center for Security and Emerging Technology and previously served on Taiwan’s National Security Council.
Sam Bresnick is a research fellow and an Andrew W. Marshall fellow at Georgetown University’s Center for Security and Emerging Technology.
