China’s post-Mao attempt to balance total control of thought and expression in many areas of work and life with circumscribed freedom of mind in certain scientific and technological areas has yielded dividends, suggests Walter Donway. But the policy has definite limits (The Daily Economy, March 16, 2026).
Indiscriminate, massive assaults on everyone and everything all the time, from food production to culture to simple existence, as inflicted by the Maoist policies of the Great Leap Forward (1958-1962) and the Cultural Revolution (1966-1976) are necessarily more destructive of life and means of living than more selective massive assaults on people and endeavors. If you live in China, the latter is better for you—unless you belong to the wrong group.
Donway writes:
China’s true ascent in AI began not at home but abroad. In the 1980s and 1990s, waves of Chinese students were sent to study at Western universities—especially in the United States. Tens of thousands entered programs in electrical engineering, computer science, applied mathematics, robotics, and cognitive science. This cohort would become the seedbed of China’s scientific and technological elite.
The same regime that had destroyed China’s intellectual class now sought to rebuild it—but to rebuild it inside a cage. Science was to be liberated—up to a point. The mind could be free in the laboratory if it served national rejuvenation, but not in other realms—above all, not political thought.
Thus was born what might be called the principle of segmented freedom: autonomy in STEM, obedience in everything else….
What made [an institution like Microsoft Research Asia founded by Chinese who had studied abroad] extraordinary—and emblematic of the Chinese model—was its borrowed freedom. It operated with an autonomy unmatched in other sectors of China’s intellectual life. Political discussion remained off-limits, but scientific inquiry was encouraged, even celebrated. The state tolerated this exceptional zone of independence because it served a higher political objective: national technological power….
Foundational breakthroughs [in artificial intelligence]—those requiring leaps of conceptual imagination—continued to come disproportionately from the West. China’s strengths lay overwhelmingly in applied AI. Yet many of the deep architectural revolutions of modern AI—transformers, diffusion models, deep reinforcement learning—were developed elsewhere. When OpenAI, DeepMind, or Google introduced a paradigm shift, Chinese firms adapted and scaled it with astonishing speed, but the original leaps of abstraction were less common….
The mind seeks expression wherever it can find room to breathe. Even in unfree systems, it carves out local zones of competence, mastery, and ingenuity.
But the ceiling is real as well. Innovation under authoritarianism is conditional: adaptive but ultimately bounded. A society may import techniques created in free cultures, scale them with discipline and data, deploy them by centralized command. It may even tolerate islands of scientific autonomy so long as they serve national power. What it cannot indefinitely command are the wellsprings of innovation: the indivisible freedom of the mind to question all premises, raise all doubts, discard orthodoxies, and pursue truth without a political price tag….
The mind committed to recognizing reality as an absolute does not have “no go” zones. Habits of obedience, once learned, migrate.
Emulating and applying what creative originators come up with is common in almost any society. Not quite indicated in Donway’s perceptive discussion is the extent to which the Chinese government and Chinese industry have stolen the intellectual property of originators who are more likely to emerge and more able to function in the most free societies. Among the beneficiaries of the systematic pillage are China’s defense industry, communications industry, AI products like DeepSeek, and electric cars.
Ergo, supplement Donway’s essay with Shedd and Badger’s book The Great Heist: China’s Epic Campaign to Steal America’s Secrets. The authors begin with the story of how, in 2018, just months after Elon Musk closed a deal to make and sell Teslas in China, a soon-to-be-former Tesla engineer named Guangzhi Cao stole the source code “for Autopilot, the advanced driver assistance system that allows Tesla vehicles to drive automatically…. Then came the real bombshell: Cao was going to work for XMotors [Xpeng], a Chinese electric vehicle startup backed by powerful interests and flush with Chinese government money.”
Also see:
StoptheCCP.org: “China’s Theft of IP Is Only Part of the Problem”
