Despite the incredibly racist and difficult environment in California for American Chinese, they went to war for their country.

Though the memories are hazy after all this time, Margery Wong remembers 1944 like it was yesterday. When news came to her front door that her brother, then 20-year old, Sergeant Yuen Hop, was missing in action.

“He enlisted when he was about 18. I was probably about 12 years old. My dad was working in the orchards…and my mom…I think she took it pretty bad.”

Army Air Force Sgt. Yuen Hop’s plane had been shot down on a mission in Germany. Details at the time were slim.

[…]

About 20,000 Chinese Americans served in World War II, even despite the Chinese Exclusion Act.

“America First” (Nazi Americans) tried extremely hard to prevent these American Chinese from becoming citizens, and yet they still served with distinction and gave the ultimate sacrifice for America (killed by Nazi Germans).

As a student of information warfare and American history, I can’t help but notice the unsettling parallels between today’s federal workforce crisis and the Bonus Army situation of 1932. In both cases, we see a fundamental conflict over promised government compensation during times of economic uncertainty.

Nearly 100 years ago American military veterans marched en masse on Washington demanding early payment of service certificates they were owed.

Just before Congress adjourned in the summer of 1932, thousands of desperate World War I veterans surrounded the U.S. Capitol. With the nation in the grips of the Great Depression, the House of Representatives had approved a bill to provide immediate cash payments to veterans. Servicemembers now waited anxiously as the Senate debated the same bill. At issue was the question, What did the nation owe its veterans?

Notably, servicemember camps setup around the Capitol were “racially integrated, vibrant communities“, a very alarming situation to those holding a line on extremely racist power. In other words the outspoken primary opponent to giving veterans money owed was from Senator Joseph T. Robinson of Arkansas – a violent segregationist during the Jim Crow era who also opposed anti-lynching legislation. He rose to Senate Majority Leader in 1933 where he blocked all civil rights legislation for the next four years.

The Hoover administration’s response to the integrated Americans demanding rights was to characterize military veterans – who had served their country faithfully and had followed the proper channels for basic income – as opportunists and troublemakers. Today, we hear similar rhetoric, with the White House describing federal workers as “lazy” and accusing them of “ripping off the American people” by doing their jobs.

The Bonus Army crisis escalated when the Hoover administration sent the racist and segregationist General Douglas MacArthur to forcibly remove the veterans from their integrated encampments. For context, MacArthur’s tenure as Army Chief of Staff from 1930 to 1935 heavily promoted institutional racism. Later, as Supreme Commander of Allied Powers in occupied Japan, he again promoted racial segregation policies. However it was his conduct during the Korean War that has drawn the most scrutiny from historians in terms of his deeply flawed strategic decision making – racism against Asians caused unnecessary losses due to consistent underestimation of Chinese military capabilities.

Thurgood Marshall recalled that General MacArthur, who believed that American Blacks were inferior to whites, was the greatest impediment to the Army’s desegregation in Korea. Things changed rapidly as soon as Truman fired him in 1951.

Today’s situation, while different in its details, shows similar signs of escalating tension, as well as racist underpinnings to attacks on diverse groups of federal workers. The administration’s recent memo threatening those who remain with “enhanced standards of suitability and conduct” and warning of prioritized “investigation and discipline” creates an atmosphere of intimidation reminiscent of the Jim Crow era.

Just as the Bonus Army veterans faced uncertainty about whether they would ever receive their promised compensation, today’s federal workers are being offered a deal that unions warn may never be paid because it lacks congressional authorization. The administration’s pressure tactics – including warnings about impending layoffs and demands for “loyalty” – echo the kind of strong-arm approaches that characterized the government’s response to the Bonus Army.

What’s particularly concerning is how this situation could potentially escalate. The Bonus Army crisis became a watershed moment in American history not because of the initial dispute, but because of how the government chose to handle it. The sight of American troops attacking American veterans created a public relations disaster that contributed to Hoover’s defeat.

Today, we’re seeing scattered protests outside federal buildings. One worker quoted in recent reporting expressed fear that “we’re all going to lose our jobs and they’re going to put all these loyalists or people that will be their shock troops.” This language of “shock troops” and loyalty tests yet again eerily mirrors the militaristic response to the Bonus Army.

The critical difference now is that we have the benefit of historical hindsight.

The Bonus Army crisis teaches us that handling of regular government workers – whether veterans or civil servants – as enemies of the state rather than as dedicated public servants tends to backfire both politically and practically. When the racist segregationist MacArthur led troops against the desegregated Bonus Army, he wasn’t just attacking a group of protesters – he was attacking the very idea that the government should honor its commitments to all those Americans who serve it.

In the current situation, federal courts have already stepped in to temporarily halt the administration’s plans. This judicial intervention offers hope that we might avoid the kind of confrontation that marked the Bonus Army crisis. However, the administration’s rhetoric and tactics suggest they may not be learning the correct lessons that history offers.

The ultimate resolution of the Bonus Army crisis – Congress finally authorizing early payment in 1936 – reminds us that these situations eventually require legislative solutions, not executive force. Today’s federal workers, like the veterans of 1932, are simply asking the government to treat them with the respect and consideration they’ve earned through their service.

As we watch this situation unfold, we would do well to remember that the Bonus Army crisis didn’t have to end in tear gas and burning encampments. It escalated because racist leadership chose confrontation over negotiation, forceful bluster over competency in dialogue. Let’s hope today’s leaders can learn from clear historical mistakes before we witness another awful MacArthur moment in American history.

Trump keeps praising a controversial American general whose actions nearly prompted World War III: “MacArthur was considered a ‘media whore’ of his time, Daniel Drezner, a professor of international affairs at Tufts University, told Reuters.” […] “I fired him because he wouldn’t respect the authority of the president,” Truman later explained. “I didn’t fire him because he was a dumb son of a bitch, although he was, but that’s not against the laws for generals.”

Truman later said that he had become anti-racist by 1946, which perhaps helps explains why he put such an abrupt end to huge “dumb” mistakes being made by an obviously racist MacArthur. And also explains why Trump keeps trying to repeat those same mistakes.

Xenophobic Political Theater Undermines Security and Innovation

Recent legislation proposing to “decouple” American AI from China presents itself as a national security measure. However, it follows a disturbing historical pattern of technological nationalism that has repeatedly harmed both innovation and human dignity in American history.

The rhetoric surrounding this AI legislation eerily mirrors the Chinese Exclusion Act of 1882, the first major law restricting immigration based on nationality.

Both share a fundamental contradiction: wanting Chinese labor and expertise while simultaneously seeking to exclude Chinese people and contributions.

Stanford University’s own history exemplifies this contradiction. Chinese laborers were instrumental in building the transcontinental railroad that created the fortune of Leland Stanford. These workers faced brutal conditions, earned lower wages than their white counterparts, and were forced to work in the most dangerous conditions – leading to the horrific practice of measuring tunnel-blasting time as a “Chinaman minute,” calculating how many Chinese would be killed per unit of progress.

Yet after the railroads were built, these same workers were blocked from citizenship, property ownership, and economic advancement through systematic legal discrimination. Stanford University itself, built with wealth from Chinese labor, would later enforce strict quotas on Chinese students.

Notably the proposed AI bill literally suggests 20 years of jail for anyone found using a computer that has evidence of “models” from China.

According to the language of the proposed bill, people who download AI models from China could face up to 20 years in jail, a million dollar fine, or both.

Imagine the kind of politician who dreams of a future where simply dropping a model on someone’s computer gets them swept off to jail for decades.

This pattern of wanting Chinese contribution while seeking to criminalize and exclude Chinese influence is technically impossible in modern AI development. The reality of artificial intelligence creation defies such artificial boundaries. Modern AI stands on the shoulders of Chinese-American innovators, perhaps most notably in the case of ImageNet. This revolutionary dataset, which transformed computer vision and helped launch the deep learning revolution, was created by Fei-Fei Li, who immigrated from China to the US and led the project at Stanford. The story of ImageNet exemplifies how arbitrary and harmful national divisions in technology can be.

The technical integration goes even deeper when we examine a typical large language model. Its attention mechanism derives from Google’s transformer architecture, developed in the US, while incorporating optimization techniques pioneered at Tsinghua University. The model runs on GPUs designed in America but manufactured in Taiwan, implementing deep learning principles advanced through international collaboration between researchers like Yann LeCun in the US and Jian Sun at Microsoft Research Asia. The training data necessarily incorporates Chinese web content and academic research, while the underlying software libraries include crucial contributions from developers worldwide, including significant optimizations from teams at Alibaba and Baidu.

The recent panic over DeepSeek perfectly illustrates this contradiction. While lawmakers rush to ban a Chinese AI model from American devices, they ignore that any open model can be fully air-gapped and run offline without any connection to its origin servers. Meanwhile, Zoom – which handles everything from classified research discussions to defense contractor meetings on university campuses – frequently processes data through Chinese code and infrastructure with minimal or no security review.

Or consider Tesla, whose Nazi-promoting eugenicist South African CEO openly praises China as his favorite government with the largest market and manufacturing base while simultaneously acquiring unprecedented access to U.S. government contracts and infrastructure projects. There couldn’t be a more obvious immediate national security threat than Elon Musk.

This inconsistent approach – where some Chinese technology connections trigger alarm while others are ignored – reveals how these policies are more about political theater than actual security. Real security would focus on technical capabilities and specific vulnerabilities, not simplistic national origins.

Modern AI development relies on an intricate web of globally distributed computing infrastructure. The hardware supply chains span continents, while the open-source software that powers these systems represents the collective effort of developers across the world. Research breakthroughs emerge from international collaborations, building on shared knowledge that flows across borders as freely as the data these systems require.

Let’s be clear: espionage threats are real, and data security across borders is a legitimate concern. However, the fundamental weakness of nationalist approaches to security is their reliance on binary, overly simplistic classifications. “Chinese” versus “American” AI creates exactly the kind of rigid, brittle reactive security model that sophisticated attackers exploit most easily, making everything less safe.

Robust security systems thrive on nuance and depth. They employ multiple layers of validation, contextual analysis, and sophisticated inference to detect and prevent threats. A security model that sorts researchers, code, or data into simple national categories without thought is like France thinking defense meant an expensive concrete Maginot line, without trucks or planes. No adaptive, intelligent internal security means one simple breach compromises everything. Real security requires technical solutions that can handle nuance, detect subtle patterns, and proactively adapt to emerging threats.

The defense of Ukraine against Russian invasion demonstrates this principle perfectly. When Russian forces attempted to seize Hostomel Airport near Kyiv, Ukraine’s defense succeeded not through static border fortifications, but through rapid threat identification and adaptive response. While main forces were positioned at the borders, regular troops at Hostomel identified and engaged helicopter-borne attackers immediately, holding out for two crucial hours until reinforcement brigades arrived. This adaptive defense against an unexpected vector saved Kyiv from potential capture. Similarly, effective AI security requires moving beyond simplistic border-based restrictions to develop dynamic, responsive security frameworks that can identify and counter threats regardless of their origin point.

Consider how actual security breaches typically occur: not through broad categories of static nationality, but through specific vulnerabilities, social engineering, and careful exploitation of trust boundaries. Effective security measures therefore need to focus on technical rigor, behavioral analysis, and sophisticated validation frameworks – approaches that become harder, not easier, when we artificially restrict collaboration and create simplistic trust boundaries based on nationality.

Just as the Chinese Exclusion Act failed to address real economic challenges while causing immense human suffering, this attempt at technological segregation would fail to address real security concerns while hampering innovation and perpetuating harmful nationalist narratives.

The future of AI development lies not in futile attempts at nationalistic segregation and incarceration, but in thoughtful collaboration guided by strong technical standards and security frameworks. The global nature of AI development isn’t a vulnerability to be feared, it’s a strength to be lead by embracing technical excellence and rigorous security practices that focus on capabilities rather than simplistic national origins.

As we stand at the dawn of the AI era, we face the same choice that has confronted American innovation throughout history: we can repeat the mistakes of xenophobic restrictions that ultimately harmed both American security and human dignity, or we can embrace the inherently open nature of technological progress while building the technical frameworks needed to ensure its responsible development.

In the early 1970s, two revolutionary computer networks were taking shape on opposite sides of the Americas. Both aimed to connect distant points for rapid information sharing. Both used cutting-edge technology. Both would transform how we think about communication and control. Yet their fates could not have been more different.

ARPANET, developed under the protective umbrella of the U.S. Department of Defense, would evolve into today’s internet. Project Cybersyn, Chile’s experiment in economic cybernetics, would end in flames during Pinochet’s coup. The key difference wasn’t technological – it was political legitimacy.

The Power of Military Cover

ARPANET enjoyed a crucial advantage: military backing provided both resources and legitimacy. When you’re developing technology for the Defense Department, questions about control and surveillance take on a different character. They become matters of national security rather than political power.

This military cover allowed ARPANET to develop largely shielded from political scrutiny. Its potential for surveillance and control was present from the beginning, but the military context made such capabilities appear necessary and appropriate rather than threatening.

Cybersyn’s Fatal Transparency

In contrast, Project Cybersyn wore its civilian nature openly. Stafford Beer’s vision of cybernetic management for Chile’s economy was explicit about its aims to redistribute economic control and enable worker participation. This transparency, while admirable, made it vulnerable.

The context is crucial here: Nixon’s disastrous Vice Presidential trip through Latin America in 1958 had left deep scars in U.S.-Latin American relations. Nixon himself, a toxic racist, was deeply convinced non-whites were too primitive to be capable of self-governance, even before he was elected President. When Allende’s socialist government began developing an advanced nationwide computer network for economic management, it triggered immediate alarm for Nixon. Without the protective shield of military classification, Cybersyn’s capabilities for coordination and control during the Cold War were seen as purely political threats.

The Hidden Legacy of Wartime Cybernetics

What’s often overlooked is how both projects drew from the same well: wartime developments in operations research and cybernetics. The core ideas about network architecture, feedback loops, and distributed control had been developed during World War II for military purposes. Many of the key figures had wartime experience in these fields.

The difference was in how this military heritage was acknowledged. ARPANET maintained its explicit military connection, which paradoxically made its civilian applications appear less threatening when they emerged. Cybersyn, by transparently adapting military-derived techniques for civilian use, found itself more vulnerable to political attack.

The Pattern Continues

This dynamic – military cover providing political legitimacy for transformative technologies – wasn’t unique to ARPANET. GPS, another technology that enables unprecedented tracking and coordination, followed a similar path. By remaining under military control during its development, it avoided many of the political questions that might have surrounded a civilian positioning system.

Even today, we see this pattern continuing. Many surveillance and control technologies face less scrutiny when developed for military or national security purposes than when proposed for civilian use. The military origin serves as a kind of political shield, even as these technologies inevitably find their way into civilian applications.

Learning from History

The parallel stories of ARPANET and Cybersyn offer important lessons for understanding how transformative technologies gain social acceptance. The political legitimacy conveyed by military backing can be crucial for a technology’s survival and evolution, even when its ultimate applications are primarily civilian.

This history should prompt us to question our different reactions to similar technologies based on their institutional origins. When we celebrate ARPANET as the precursor to the internet while dismissing Cybersyn as an authoritarian fantasy, we’re not just making a technological assessment – we’re revealing deep-seated assumptions about legitimate and illegitimate forms of control.

The next time we encounter a new technology that promises to reshape social organization, we might ask ourselves: Would our reaction be different if this came from the military rather than civilian sector? The answer might tell us more about our political assumptions than about the technology itself.

The Lesson for IT Pros Today

The stories of ARPANET and Cybersyn aren’t just about the past – they’re a warning about how institutional control can be seized through technical means. When Pinochet’s forces moved against Cybersyn, they understood that controlling these networks meant controlling Chile’s infrastructure, economy, and ultimately, its people. The system’s designers and operators faced a stark choice: hand over the keys to power, or resist.

Today, American technical professionals face eerily similar choices. When demands come down to centralize access to critical systems – whether they’re financial networks, health databases, or communication infrastructure – these aren’t merely technical requests. They’re potential preludes to institutional capture. The separation between these systems isn’t bureaucratic inefficiency; it’s a deliberate firewall against concentrated power.

The technical professionals who manage these systems aren’t just administrators – they’re the last line of defense for institutional independence.

Every request for master access, every demand to break down system isolations, every push to centralize control should be viewed through the lens of Chile’s experience. The question isn’t whether such access is technically feasible – it’s whether it preserves the separations that protect democratic governance.

For those who work in American institutions today, Cybersyn’s fate offers clear lessons: Technical architecture is political architecture. System access is institutional power. And sometimes, protecting democracy means having the courage to say “no” to those who would dismantle its safeguards in the name of efficiency or security.

The choice between being a mechanism of control or a guardian of democratic institutions isn’t just a historical artifact of the Internet origin; it’s an ever-present reality.