NARRATIVE: China is stealing American AI — while Washington looks the other way

Introduction: A report that should surprise no one

CrowdStrike and the number that stings

In June 2026, the American cybersecurity firm CrowdStrike published a finding that should have dominated headlines but was eclipsed by war and diplomacy: Chinese entities accounted for more than half of all state-sponsored intrusions targeting technology companies — particularly their artificial intelligence assets — in the 12 months ending March 31, 2026. More than half. This is not a worrying trend to monitor. It is an active, massive, documented campaign of industrial technological theft.

Along the same lines, the FBI issued an unequivocal statement: "China's economic espionage campaign costs the American economy hundreds of billions of dollars per year and endangers our national security." These figures, these statements, and this data are known. What is less known is that while American intelligence agencies were documenting this threat, the U.S. Commerce Department was quietly backing away from blacklisting more than 100 Chinese companies that had already been approved by the interagency committee as security risks.

DeepSeek and the companies that were spared

The blacklist that is no longer black

The Foundation for Defense of Democracies (FDD) report from June 27, 2026 tells the following story: the American interagency committee had approved adding more than 100 Chinese companies to the Commerce Department's Entity List — a list that restricts American exports to entities posing security risks. These approvals had been granted on the basis of coordinated risk assessments by multiple intelligence agencies.

But the Commerce Department refused to proceed with the blacklistings in order not to provoke a new escalation with Beijing. Among the companies spared was DeepSeek, the Chinese AI startup whose R1 model had raised specific security concerns. This diplomatic choice — forgoing security measures to avoid commercial friction — illustrates exactly the contradiction at the heart of American policy toward China in 2026.

Why DeepSeek deserves attention

DeepSeek is not a simple technology company. Its R1 model, presented as an open-source competitor to OpenAI's models, sent shockwaves through the American tech sector in early 2026. Its ability to rival leading American models at a fraction of the declared cost raises serious questions about its development methods — notably the possibility that this development benefited from data or technologies obtained through illegal channels.

The fact that the interagency committee recommended placing DeepSeek on the Entity List, and that this recommendation was overridden for diplomatic reasons, is revealing. It says that short-term economic interests with China outweighed documented national security concerns. This is not a trivial decision in a world where the technological competition is also a military security issue.

China's economic espionage strategy

Systematic targeting of AI assets

CrowdStrike's data reveals specific and strategic targeting: technology companies whose artificial intelligence assets are the primary objective. This is not opportunistic espionage — it is a coordinated national policy aimed at closing the technological gap between China and the United States in the field of AI. Beijing has understood that the next wave of military and economic power will be shaped by AI — and that it is faster and less costly to steal than to develop.

This targeting takes several documented forms. Direct intrusions into companies' computer systems. "Insider" risks — employees or contractors who deliberately transmit information to Chinese actors. Recruitment via professional platforms such as LinkedIn and Indeed, documented in a Five Eyes report cited by Japan Forward. And investments in American AI startups that provide access to technologies through legal mechanisms.

Documented access vectors

The Five Eyes report — the intelligence alliance comprising the United States, United Kingdom, Canada, Australia, and New Zealand — documented how China uses LinkedIn and Indeed to recruit Western technology professionals, either as spies or as unwitting vectors of technology transfer. These platforms, legitimately used by billions of people, become tools for approaching potential targets.

The company Qihoo 360, described by Forbes as the "cyber giant behind China's Mythos rival," illustrates another dimension of this strategy: Chinese technology companies that develop offensive capabilities within a semi-public ecosystem that blurs the lines between private sector and state operations. This model — surveillance capitalism in the service of state espionage — is one of the asymmetric advantages China exploits in its competition with the West.

The connection to Ukraine: China fuels Russia

Technological theft in the service of war

There is a direct link between Chinese economic espionage in the field of AI and the war in Ukraine — and it deserves to be named explicitly. China is stealing artificial intelligence technologies that have direct military applications: autonomous guidance systems, image recognition for drones, optimization of military supply chains. Once mastered, these technologies can be shared or sold to Russia, which uses them in its drones and missile guidance systems.

This indirect technology transfer — from American industry, via Chinese espionage, to the Russian military — is difficult to document precisely. But its existence is plausible enough that American intelligence agencies have identified it as a strategic risk. The decision to spare DeepSeek and 100 other Chinese companies from the Entity List must be read in this context: it potentially facilitates a technological flow that indirectly feeds the war against Ukraine.

Beijing as enabler of the Russian war machine

Beijing has always denied supplying weapons directly to Russia. But the definition of "military assistance" is far broader than arms deliveries. Maintaining commercial flows with Russia under sanctions, providing dual-use (civilian and military) components, and the potential technological transfer in the fields of AI and drones constitute substantial support to the Russian war machine — even if no Chinese tank has been delivered to the Ukrainian front.

This reality is explicitly acknowledged by Western governments in their statements, even as they often hesitate to draw the consequences in terms of policy toward Beijing. The decision to spare Chinese companies from the Entity List in the name of diplomatic appeasement compounds this contradiction: we know what China is doing, we name the problem, and we choose not to act on it sufficiently.

The American response: between sanctions and concessions

The architecture of export controls

The United States has a legal arsenal to counter Chinese economic and technological espionage. The Commerce Department's Entity List is one of its primary instruments. Export controls on advanced semiconductors, including restrictions imposed on NVIDIA and its partners, aim to deny China the chips needed to train cutting-edge AI models. However, the Trump government's decision to withhold its most advanced AI models — Mythos and Fable developed by Anthropic — from allies creates new friction: it slows technology sharing with reliable partners while doing nothing to prevent theft by adversaries.

The coherence of this policy is questionable. Advanced technologies are restricted from allies for reasons of control. Chinese companies identified as security risks are spared for diplomatic reasons. Cybersecurity programs are funded while recruitment platforms used by Chinese intelligence services operate freely. These contradictions are not the doing of any specific administration — they reflect the structural difficulty of simultaneously managing economic, diplomatic, and security interests that often conflict.

What allies expect from Washington

For American allies in Europe and Asia, the central question is not whether China is spying. That is a documented certainty. The question is whether Washington will coordinate a collective, coherent response — shared with its allies — or continue managing this issue unilaterally and inconsistently. The Five Eyes have produced a joint report on China's recruitment via LinkedIn — that is an example of working coordination. It needs to be expanded.

The NATO summit in Ankara is primarily focused on Russia and Ukraine. But China is there too — in the background, in discussions about technological dependency, defense supply chains, and cybersecurity capabilities. The twenty European conservative parliamentarians who made the trip to Washington raised these topics with the American administration. That is no coincidence.

International cooperation: the Five Eyes and the need for a coordinated response

A joint report, a shared threat

The Five Eyes report published in 2026 on spy recruitment via LinkedIn and Indeed marks an important milestone in Western intelligence coordination. For the first time, the five agencies — American (NSA/FBI), British (GCHQ), Canadian (CSE), Australian (ASD), and New Zealand (GCSB) — produced a joint warning explicitly addressed to private companies about the recruitment methods used by Chinese intelligence services. This report documents how professionally legitimate-looking profiles are used to approach researchers, engineers, and employees at sensitive technology companies.

The value of this kind of coordination extends beyond simple data aggregation. It sends a signal to Chinese espionage services that their methods have been identified. It raises awareness among potential targets in private companies. And it sets a precedent for coordinated sector-specific warnings — a model that could be expanded to NATO partners and G7 countries.

The limits of the multilateral response

Intelligence coordination has structural limits, however. Information sharing between allies is governed by strict rules on protecting sources and methods. The decision to publish a report on Chinese espionage methods means revealing part of what is known — and thus potentially compromising certain sources. This difficult calculation explains why public reports often lag behind what agencies know in real time.

There is also a political dimension. China responds strongly to any public accusation of espionage, dismissing it as "defamation" and "Cold War politics." For countries like Canada, Australia, or European nations that maintain significant economic ties with Beijing, the decision to co-sign an accusatory report carries a measurable diplomatic cost. Participation in these reports is therefore a political act, not merely an intelligence act.

The AI dimension: why this technological war is existential

AI as a military force multiplier

Artificial intelligence is not an abstract technological subject. It has concrete military implications that are transforming current and future wars. Drones equipped with autonomous target identification capabilities. Image recognition systems for missile guidance. Logistics optimization algorithms that allow armies to move more efficiently. China is investing massively in all of these applications — and according to CrowdStrike's data, it is accelerating this development by stealing research from American technology companies.

CrowdStrike's finding that Chinese entities are responsible for more than half of intrusions targeting technology companies' AI assets is not an approximate estimate — it is the result of continuous monitoring of the advanced persistent threats (APT) that characterize state espionage operations. These operations are sophisticated, patient, and long-term. They do not seek to cause immediate damage — they seek to transfer knowledge silently.

The risk for future competition

The long-term stakes are those of global technological competition. If China succeeds in closing the gap in AI technologies through a combination of internal development and technological theft, it will have military capabilities in 10 to 15 years comparable or superior to those of the United States. The implications for the security of Taiwan, for freedom of navigation in the South China Sea, and for American alliances in the Asia-Pacific are considerable.

This prospect is not an alarmist hypothesis. It is documented in American intelligence assessments, in National Security Commission on AI reports, and in analyses by independent experts. What is alarming is not the threat itself — it is known. It is the insufficient political response, illustrated by the decision to spare 100 Chinese companies from the Entity List for diplomatic reasons.

The human cost of technological espionage: what the numbers erase

Companies, researchers, careers

Behind CrowdStrike's numbers — 50% of state-sponsored intrusions — there are people. Engineers whose years of work are copied in a matter of hours. Startups whose algorithms developed over months end up integrated into competing products, with no legal recourse. Researchers approached via LinkedIn with enticing job offers, some of whom do not immediately realize they are participating in an organized technology transfer.

The FBI estimates losses at "hundreds of billions of dollars per year." This figure is real, but it tends to dematerialize the human consequence. For an AI startup that discovers its technology has been stolen, there is often no recourse. Proving the origin of an intrusion, linking stolen code to a competing product, convincing a court of the harm's existence — it is a legal battle that few companies have the resources to fight.

The culture of secrecy and its paradoxes

The institutional response to Chinese economic espionage suffers from a paradox. To be effective, cyber defense requires sharing information about attack vectors, intrusion methods, and the actors involved. But this information sharing requires admitting that one has been compromised — something most companies are reluctant to acknowledge publicly, for fear of the consequences on their reputation, partnerships, and market valuation.

This structural silence benefits espionage actors. Every compromised company that does not share its experience allows the same method to be used against others. CrowdStrike, Mandiant, and other cybersecurity firms play a crucial role in aggregating this data anonymously — but anonymous aggregation has its limits. The lack of transparency in the technology ecosystem is a systemic flaw that espionage services exploit with consistency.

Conclusion: When commercial diplomacy sabotages national security

The central contradiction

The narrative this article documents is that of a major contradiction in American policy toward China in 2026: intelligence agencies identify companies that pose security risks, propose concrete measures, and those measures are blocked by diplomatic and commercial considerations. This pattern is not new — it has characterized the American-Chinese relationship for decades. But in the context of the competition over AI and a war in Ukraine where stolen technologies could end up in Russian missiles, this pattern has consequences that extend beyond bilateral Sino-American relations.

I am not an expert in the intricacies of American trade policy. I cannot demonstrate that the 100 Chinese companies spared actually transferred technologies to Russia. But I can note that the available data — CrowdStrike, FBI, Five Eyes, FDD — converge on the same diagnosis: China is conducting a systematic campaign of technological theft in the field of AI, and this campaign is being insufficiently countered. That finding deserves to be stated clearly.

What the West must decide

The West must decide whether it treats China as a trading partner with practices to correct, or as a rival power using commerce as a lever of influence. These two visions are not easily reconciled. The first implies continuous diplomatic concessions. The second implies real economic costs — supply chain disruption, higher prices for certain goods, diplomatic tensions.

These costs are real. They are also comparable to the costs of doing nothing: a China equipped with superior military AI capabilities in 15 years, American technologies stolen for use in enemy weapons systems, and an economy whose most dynamic sectors are systematically pillaged. That calculation is not hard to make. It is hard to accept politically. That is where the real problem lies.

By Maxime Marquette, columnist

Columnist's transparency note

What I know and what I infer

This narrative rests on verifiable facts: CrowdStrike's data on Chinese intrusions (June 2026), the FBI statement cited by CNBC, the FDD report on companies spared from the Entity List, and the Five Eyes report on LinkedIn recruitment. The link between Chinese technological espionage and the war in Ukraine is a cautious inference based on documented connections — not a factual certainty. I state this explicitly.

Biases and editorial position

I regard China as the greatest long-term threat to the West — a position I hold deliberately, and one that influences the angle of this text. I am pro-Ukraine and pro-NATO. These biases do not create the facts — they influence their framing. The reader is invited to consult the primary sources directly to form their own judgment.

Sources

Primary sources

CNBC — China-linked actors target far more than technology in an intense AI race with the U.S. — July 1, 2026

FDD Action — National Security Roundup: China and company blacklist — June 27, 2026

IANS Live — US warns China is stealing AI secrets — June 27, 2026

Secondary sources

AI Weekly — Chinese espionage and AI assets — July 2026

Japan Forward — Five Eyes report: China recruiting spies via LinkedIn and Indeed — July 2026

Forbes — Qihoo 360, the cyber giant behind China's Mythos rival — June 30, 2026