America’s AI debate is a hot mess. Policymakers, technologists, investors, and activists cannot agree on the destination, the strategy, the acceptable costs, or even what “winning” the artificial intelligence race means.

That confusion shuts ordinary Americans out of a debate that will shape their jobs, businesses, institutions, and way of life. Yet an America-first AI regime cannot be designed only by Washington policymakers, Silicon Valley moguls, or a priesthood of engineers. It must serve the people who build businesses, sign paychecks, raise families, and sustain the country.

America’s technological order should be built around productive citizens rather than a handful of companies pursuing a woo-woo race to turn the planet into a cyber-utopia.

My Independence Day forecast looks more accurate now than it did when I wrote it: No single political authority controls America’s technological future, and no shared moral authority guides it. We still have the Declaration, the Constitution, federalism, and one another. Almost everything else remains unsettled.

The Kimi shock

Since July 4, factional warfare among the AI elite has intensified after the release of Kimi, a Chinese open-source model that reportedly outperformed leading American closed-source systems on important benchmarks.

Kimi did not emerge from an isolated Chinese laboratory. Its engineers trained it by distilling capabilities from American proprietary models — the very systems OpenAI, Anthropic, and their allies argue must remain closed to preserve America’s lead.

The result presents a serious problem. America has poured immense financial, scientific, and political capital into closed frontier labs. If Chinese developers can cheaply reproduce much of their practical value, Washington faces a nightmarish choice: either cede global AI leadership and control to Beijing or allow the pursuit of AI dominance to transform our form of government into something we may not recognize as “American.”

America’s historic model combines private innovation, free markets, and strategic federal investment. That model has supported both prosperity and national power. But if massive collaboration between government and closed-source hyperscalers cannot sustain an AI edge, Americans deserve a debate about what comes next — not a panic-driven transfer of more power and money to the same institutions.

How can America face the future if its greatest strategic advantage has become a liability? Without public agreement on what should replace the old model — or even what kind of country a “refounded” United States should be — what can we do next?

RELATED: The question of the AI age: How much humanism is too much?

Moor Studio/iStock/Getty Images

Three rival factions

Right now, among the AI elite, three factions have emerged in response to this problem.

The first faction argues that Kimi changes less than it appears. Chinese engineers did not surpass American frontier models across the board; they optimized around the most expensive benchmarks and competed where open systems could win. Closed labs can adjust, Washington can revise its strategy, and American dominance can survive.

That may be partly right. Frontier labs still possess enormous resources, talent, and political support. They will not disappear simply because open models become more useful.

The second faction believes such economic and political concerns will soon become obsolete. Its members expect recursively self-improving artificial superintelligence to transform or absorb human civilization. Their task, as they see it, is to accelerate the transition and ensure that it occurs under acceptable management. If done right, we would enter a new age that would be pleasurable beyond comprehension.

The test of an America-first AI policy is whether advanced technology strengthens the constitutional nation and the people who give it life.

The left-coded version appears in Effective Altruism. The right-coded version insists that America, not China, must command the coming superintelligence. Both treat maximum acceleration as a moral imperative. From that perspective, open-source AI is dangerous because it disperses capability, slows the drive toward a single frontier breakthrough, and weakens the control of the institutions leading the race.

The third faction rejects the bargain behind closed-source supremacy. Its argument is that hyperscalers are frightening Americans into financing their rise as private governors of technological progress. The companies gain ungodly wealth and authority; Washington gains the hope of planetary AI dominance; citizens inherit the cost and the risk.

Kimi strengthens that argument because it competed by optimizing for what businesses actually need: efficient computing, useful workflows, enterprise deployment, and practical web applications. Open models can undercut closed systems precisely where American economic life is strongest — decentralized experimentation by workers, entrepreneurs, and firms.

This faction therefore urges America to embrace and encourage domestic open-source AI. The point is not that every model must be public or that frontier capabilities require no safeguards. It is that America’s technological order should be built around productive citizens rather than a handful of companies pursuing a woo-woo race to turn the planet into a cyber-utopia.

RELATED: America needs an artificial intelligence coalition, not an AI fortress

Burcu Demir/iStock/Getty Images

The real America-first path

Of the three positions, the third offers the clearest route to America-first AI because it begins with the right political subject: the American people. Technology should enlarge their capacity to create, compete, govern themselves, and live well. They should not be treated as raw material for a government-corporate project aimed at controlling the future.

The first faction is also correct that closed frontier labs will remain part of the system. Some Americans genuinely want to invest heavily in the most powerful possible models. The federal government also believes its military, financial, and geopolitical position may depend on sovereign access to frontier AI.

That belief creates the greatest danger.

Washington may conclude that preserving its global dominance requires technological control beyond civilian resistance. If officials decide they must choose between the survival of the federal regime and the survival of America’s constitutional order, “national security” will become the justification for hoarding data, compute, models, and authority on a scale the country has never seen.

China would then win without beating America at the frontier. By flooding the world with useful, inexpensive, business-friendly open models, Beijing could tempt Washington into becoming more centralized, secretive, and coercive — less recognizably American.

America should refuse that trap. It should protect genuinely dangerous capabilities, preserve a competitive frontier sector, and make room for open models that empower American businesses and citizens. The test of an America-first AI policy is not whether it produces the most powerful machine at any cost. The test is whether advanced technology strengthens the constitutional nation and the people who give it life.

If our AI strategy requires America to stop being America, China has already won.