China has penetrated Taiwan's military. Does that make US intelligence sharing dangerous?
Taiwan charged 64 people with espionage in 2024, most of them military personnel. As Washington deepens intelligence cooperation with Taipei, every secret shared risks reaching Beijing. The question is whether the deterrent value exceeds the counterintelligence cost.
🎧 Listen to this article
The Leak That Keeps on Giving
Taiwan’s Ministry of National Defense has a problem it cannot solve by firing spies. In 2024, the island’s National Security Bureau charged 64 individuals with espionage—two-thirds of them current or retired military personnel. China’s recruitment methods are blunt: money, debt leverage, internet cultivation, and the patient courting of retired officers who retain access to old colleagues. The penetration is not hypothetical. It is documented, ongoing, and structural.
This creates an uncomfortable question for Washington. The United States shares intelligence with Taiwan to strengthen deterrence. But if Beijing has compromised Taiwan’s command structure, every piece of American intelligence that flows to Taipei risks flowing onward to the People’s Liberation Army. In a crisis—a blockade, a missile barrage, an invasion—the very act of helping Taiwan could help China more.
The calculus is not simple. Intelligence sharing is not a tap that can be turned off without consequence. Taiwan’s ability to defend itself depends on knowing what China is doing, when, and where. American satellites, signals intercepts, and analytical products fill gaps that Taiwan cannot fill alone. Cutting the flow would leave Taiwan blind precisely when it needs to see. Yet continuing the flow means accepting that some fraction of shared intelligence will reach Beijing.
This is not a hypothetical dilemma for academic conferences. Taiwan’s 2025 Quadrennial Defence Review explicitly calls for deeper intelligence exchanges with the United States, including high-level strategic dialogues and joint tabletop exercises. Washington must decide how much to share, knowing that the recipient is compromised.
The Architecture of Compromise
Understanding the liability requires understanding the architecture. American intelligence reaches Taiwan through the American Institute in Taiwan (AIT), a nominally private organisation established under the Taiwan Relations Act to maintain unofficial relations. The arrangement is a diplomatic fiction—everyone knows AIT functions as an embassy—but the fiction matters because it preserves ambiguity about America’s commitment.
Intelligence sharing operates within this ambiguity. The United States does not have a formal alliance with Taiwan. It has no treaty obligation to defend the island. What it has is a statutory commitment to provide defensive articles and services, and a deliberate refusal to specify what happens if China attacks. This strategic ambiguity is meant to deter both Beijing and Taipei: China cannot be certain America will stay out; Taiwan cannot be certain America will come in.
The problem is that ambiguity complicates information security. Formal allies like the Five Eyes nations operate under shared classification systems, joint counterintelligence protocols, and decades of institutional trust. Taiwan operates outside this architecture. Intelligence Community Directive 403 governs disclosure to foreign entities and mandates protection of sources and methods—but Taiwan’s unique status means the normal frameworks do not cleanly apply.
What actually gets shared? The public record suggests a mix: strategic assessments of PLA capabilities, satellite imagery of Chinese military movements, and analytical products that help Taiwan understand the threat environment. The more sensitive the intelligence, the more restrictive the handling. But even carefully compartmented information can leak if the recipient’s command structure is compromised.
China’s penetration tactics target precisely the nodes where American intelligence would flow. According to Taiwan’s National Security Bureau, Beijing employs four infiltration tactics: recruiting active personnel through retired contacts, building networks via the internet, financial inducement, and debt coercion. The PLA focuses on younger, technically skilled personnel in joint operations centres and information units—exactly where US-origin data streams would be processed.
The result is a structural vulnerability that no amount of vetting can eliminate. Taiwan can catch spies. It cannot catch them all. And it cannot know which ones it has missed.
The Paradox of Useful Intelligence
The question is not whether to share intelligence with Taiwan. The question is what kind of intelligence, through what channels, with what safeguards—and whether the safeguards can ever be sufficient.
Consider the spectrum of intelligence products. At one end: strategic assessments of Chinese capabilities and intentions. These are valuable but perishable. If Beijing learns that Washington believes the PLA cannot sustain an amphibious assault for more than 72 hours, Beijing can adjust its planning. The intelligence becomes less useful to Taiwan and potentially useful to China. But the damage is limited because strategic assessments are inherently provisional.
At the other end: real-time tactical intelligence during a crisis. Satellite feeds showing Chinese naval movements. Signals intercepts revealing PLA command decisions. Targeting data for Taiwan’s missile batteries. This intelligence is extraordinarily valuable—and extraordinarily dangerous if compromised. If China knows what Taiwan knows, it can deceive, misdirect, and exploit. Worse, it can identify and neutralise the sources of American collection.
The conventional response is compartmentalisation. Share less. Restrict access. Create tiered disclosure levels. But compartmentalisation has costs. Intelligence that cannot be acted upon is intelligence wasted. Taiwan’s military needs to integrate American products into its own planning, training, and operations. Every layer of restriction reduces the intelligence’s utility.
There is also the problem of what security researchers call “diff analysis.” Taiwan’s command structure generates its own intelligence—reports, assessments, communications. If China already has access to this baseline, American intelligence becomes visible by contrast. The US tries to strengthen Taiwan’s picture by adding detail, but this creates a cleaner signal for PLA collectors who already own the base text. The enhancement becomes the tell.
The US-China Economic and Security Review Commission noted in 2025 that China is “rapidly advancing toward its goal of being prepared to take Taiwan by force,” with new amphibious assault ships and mobile piers. This is precisely the kind of assessment that helps Taiwan prepare—and precisely the kind that becomes less useful if Beijing knows Taiwan has it.
When Crisis Concentrates the Mind
The liability sharpens in a crisis. Peacetime intelligence sharing is bad enough if compromised, but the stakes are manageable. Wartime intelligence sharing is another matter entirely.
Imagine a blockade scenario. Chinese naval forces encircle Taiwan, interdicting commercial shipping. Taiwan needs to know where the gaps are, which Chinese vessels are where, what the PLA’s rules of engagement permit. American surveillance assets can provide this. But if the information reaches Beijing as quickly as it reaches Taipei, China can close the gaps, reposition forces, and exploit Taiwan’s predictable responses.
Or consider a missile campaign. China launches precision strikes against Taiwan’s command nodes, air defences, and communications infrastructure. Taiwan’s survival depends on dispersal, deception, and rapid adaptation. American intelligence on incoming salvos could save lives—but only if China doesn’t know that Taiwan knows. If Beijing can monitor what Washington shares, it can time its strikes to catch Taiwan acting on compromised information.
The deepest danger is source compromise. Intelligence agencies protect sources and methods with religious intensity because losing them is catastrophic. A human source inside the PLA, once identified, is dead or imprisoned. A signals collection method, once known, is countered. A satellite’s orbital parameters, once exposed, can be tracked and blinded. If sharing intelligence with Taiwan means sharing it with China, the cost is not just the intelligence itself but the entire collection architecture that produced it.
This is why the most sensitive American intelligence may never reach Taiwan at all. The NOFORN marking—No Foreign Nationals—exists precisely to prevent this kind of leakage. But NOFORN intelligence that could save Taiwan is intelligence that Taiwan cannot use. The marking creates what one analyst called a “Schrödinger-like state where intelligence simultaneously exists and doesn’t exist for Taiwan recipients.”
The Uncomfortable Arithmetic
How should Washington weigh these risks? The honest answer is that no formula exists. The variables are too uncertain, the scenarios too various, the stakes too high for algorithmic optimisation.
But some principles emerge. First, the value of intelligence sharing depends on Taiwan’s ability to act on it. If Taiwan’s command structure is so compromised that operational planning leaks to Beijing, tactical intelligence becomes counterproductive. Strategic intelligence retains some value because it shapes long-term preparation, but even this diminishes if China can adjust its posture in response.
Second, the risk of sharing depends on the speed of compromise. If Chinese penetration operates on a lag—weeks or months to extract and analyse—then time-sensitive intelligence may retain utility. Taiwan can act before Beijing can react. But if penetration is real-time, through cyber intrusion into Taiwan’s networks or human sources with immediate access, the window closes.
Third, the alternative to sharing is not zero risk. If the United States withholds intelligence and Taiwan is defeated, the strategic consequences dwarf any counterintelligence losses. A Chinese-controlled Taiwan transforms the Western Pacific. American credibility with every ally evaporates. The question is not whether to accept risk but which risks to accept.
The Jamestown Foundation’s analysis of PRC infiltration cases suggests that China’s intelligence apparatus is patient and systematic. It does not need every secret—it needs enough secrets to understand American and Taiwanese intentions, capabilities, and vulnerabilities. Partial penetration may be sufficient for Beijing’s purposes.
This means that even aggressive counterintelligence cannot eliminate the liability. Taiwan can reduce it. Taiwan can complicate Chinese collection. But Taiwan cannot achieve the security baseline that Five Eyes partners maintain through decades of institutional integration.
What Could Change
Three intervention points exist, each with trade-offs Washington must weigh honestly.
Tiered disclosure with degradation protocols. The United States could establish explicit tiers of intelligence sharing, with automatic degradation as crisis intensity increases. Peacetime: broad strategic assessments. Elevated tensions: selected operational intelligence with restricted distribution. Active conflict: only what Taiwan needs in the next 24 hours, delivered through channels that bypass compromised nodes. The trade-off: complexity breeds confusion, and confusion in crisis kills.
Alternative delivery architectures. If Taiwan’s command structure is compromised, route around it. American intelligence could flow directly to dispersed, hardened units rather than through centralised headquarters. Satellite communications with beam-steering and burst transmission could reduce interception risk. The trade-off: this requires infrastructure Taiwan does not have and time Taiwan may not get.
Acceptance of leakage as cost of deterrence. Washington could conclude that the deterrent value of intelligence sharing exceeds the counterintelligence cost—that a Taiwan which can fight, even imperfectly, deters better than a Taiwan left blind. This is the logic of Taiwan’s 2025 defence review, which seeks deeper exchanges despite known vulnerabilities. The trade-off: every piece of shared intelligence becomes a potential gift to Beijing.
The most likely outcome is muddling through. Washington will share some intelligence, restrict other intelligence, and hope that Taiwan’s counterintelligence improves faster than China’s penetration deepens. This is not a strategy. It is an accommodation to uncertainty.
FAQ: Key Questions Answered
Q: Has China actually penetrated Taiwan’s military, or is this speculation? A: Taiwan’s own National Security Bureau documented 64 espionage cases in 2024, with two-thirds involving current or retired military personnel. The penetration is confirmed, not hypothetical.
Q: Why can’t Taiwan simply improve its counterintelligence? A: Taiwan has increased counterintelligence efforts, but structural factors work against it. Retired personnel retain social connections to active units. Financial pressures create recruitment opportunities. And cyber intrusion bypasses human vetting entirely.
Q: Would the US really withhold intelligence if Taiwan were under attack? A: Withholding intelligence during an attack would likely doom Taiwan’s defence and devastate American credibility. The more realistic question is what kind of intelligence, through what channels, with what restrictions—not whether to share at all.
Q: How does this compare to intelligence sharing with other partners? A: Five Eyes nations operate under shared classification systems and decades of institutional trust. Taiwan lacks this foundation. The comparison is less to formal allies than to partners facing active penetration by sophisticated adversaries.
The Kintsugi Doctrine
There is a Japanese concept, kintsugi, that treats broken pottery by repairing cracks with gold. The repair is visible, even beautiful. The object is stronger for having been broken and mended.
American intelligence sharing with Taiwan cannot achieve kintsugi. The cracks cannot be gilded. But they can be acknowledged, mapped, and worked around. The alternative—pretending the cracks do not exist—is worse than accepting imperfection.
Taiwan’s command structure is compromised. This is a fact, not a failure of will. American intelligence sharing with Taiwan carries risks that cannot be eliminated, only managed. The question for Washington is not whether to accept a liability but how to minimise it while preserving the deterrent value that makes sharing worthwhile.
The answer will be imperfect. It will involve trade-offs that satisfy no one. It will require accepting that some American intelligence will reach Beijing and deciding that the cost is worth paying. This is not the clean strategic logic that briefing slides prefer. It is the messy reality of defending a partner you cannot fully trust against an adversary you cannot fully see.
Taiwan is not a problem to be solved. It is a condition to be managed. The management begins with honesty about what is known, what is compromised, and what remains worth protecting despite the leaks.
Sources & Further Reading
The analysis in this article draws on research and reporting from:
- Taiwan’s National Security Bureau 2024 Report - documentation of Chinese infiltration tactics and espionage cases
- Reuters on Taiwan’s 2025 Quadrennial Defence Review - Taiwan’s plans for deeper US intelligence cooperation
- US-China Economic and Security Review Commission 2025 Report - assessment of China’s military preparations regarding Taiwan
- Jamestown Foundation Analysis - detailed examination of PRC military infiltration cases
- Intelligence Community Directive 403 - US policy governing foreign disclosure of classified intelligence
- American Institute in Taiwan on the Taiwan Relations Act - legal framework for US-Taiwan relations
- Zero Trust Security for Coalition Operations - technical approaches to intelligence sharing with compromised partners
