Chinese labs keep releasing powerful models at rock-bottom prices. The latest, Moonshot AI’s Kimi, hit last week. It matches top offerings from OpenAI and Anthropic. Yet it costs nothing for users and carries fewer built-in limits. That single fact has set off open warfare among President Donald Trump’s current and former AI advisers.
David Sacks, who served as Trump’s AI and crypto adviser until March, fired the first shots over the weekend. He branded Anthropic’s models “lobotomized” and “woke.” He followed up by accusing leading American labs of wanting the government to “eliminate their open source competition.” His comments landed amid fresh evidence that businesses are turning to cheaper Chinese alternatives.
Emil Michael, a senior Pentagon official working with Defense Secretary Pete Hegseth, struck back hard. He labeled OpenAI’s new head of strategic futures a “supreme village idiot.” Then he turned the same insult on Dean Ball, a former Trump AI adviser now at OpenAI. Ball had criticized the White House’s new model-vetting process as a “de facto licensing regime for frontier AI.” Michael shot down any notion of quiet government pressure on companies. He insisted any action must follow “the democratic process not some Deep State scheme.”
The public barbs reveal more than personal grudges. They expose a split at the heart of U.S. AI strategy. One camp pushes for open models and minimal intervention to outpace Beijing. The other demands tighter controls, security reviews, and export limits to protect national security. And right now both sides claim Trump’s ear.
This fight comes at a bad moment for the administration. U.S. stocks already took a hit from earlier Chinese model releases. Enthusiasm for domestic AI firms has powered much of recent economic growth. When companies switch to free or low-cost options from China, that momentum fades. Anton Leicht, a fellow at the Carnegie Endowment, captured the worry in a post on X. Chinese models, he wrote, pose “a threat for an administration that really doesn’t want more economic bad news.”
Concerns run deeper than dollars. Lawmakers on both sides of the aisle have launched probes into American firms using Chinese AI. The House Committee on Homeland Security and the Select Committee on the Chinese Communist Party sent letters in April to companies including Cursor and Airbnb. They demanded details on any exposure to models from Beijing.
Andrew Garbarino, chairman of the Homeland Security panel, sounded the alarm. “The Chinese Communist Party is no longer just nipping at our heels in artificial intelligence; it is racing to close the gap in some of the exact capabilities that will shape the future of cybersecurity,” he told CNBC. He pointed to reports that open-weight Chinese models now match U.S. systems in finding software vulnerabilities.
Andy Ogles, who chairs a cybersecurity subcommittee, laid out the stakes even more bluntly. “If the cheap, capable, easy option for an AI model is Chinese, the rest of the world will build on it,” he said. “If we do nothing, Chinese models become the default foundation of the global digital economy, carrying embedded censorship, uncertain security, and capabilities distilled from our own laboratories with the safety guardrails stripped out.”
The State Department echoed those fears. A spokesperson told the same outlet that growing use of Chinese models “raises serious concerns. Those AI models are designed to advance Beijing’s narratives, censor dissent, and reflect CCP ideology and values.” Chinese officials pushed back. A spokesperson for the People’s Republic of China’s embassy in the U.K. called the charges “baseless allegations and malicious smears.” Beijing, the statement continued, builds its AI sector on “self-reliance and strength in science and technology.”
Yet the technical reality complicates easy answers. For years Washington tried to starve Chinese labs of advanced chips. Export controls tightened under President Biden. Trump loosened some of those rules. He struck a deal allowing Nvidia to sell more chips to China in exchange for a government cut. Smuggling cases still surface. Even so, Chinese teams appear to squeeze far more performance from limited hardware.
Many experts believe distillation plays a big role. Chinese models train on outputs from leading U.S. systems. OpenAI and Anthropic have complained loudly about the practice for months. In February both companies sent letters accusing firms such as DeepSeek of appropriating their technology. The Trump administration responded in April. Michael Kratsios, the president’s chief science and technology adviser, issued a memo accusing Chinese entities of “industrial-scale campaigns” to extract capabilities from American AI models.
“We will work with American companies to coordinate against China’s attacks,” Kratsios declared, according to NPR. The memo promised new detection tools, defensive measures, and potential punishments. A bipartisan House bill advanced at the same time to impose sanctions on those responsible. Beijing called the moves “unjustified suppression.”
Despite that effort, Kimi arrived anyway. It rivals the intelligence of models the U.S. government once deemed so dangerous it briefly shut down an Anthropic release for national security reasons. And it is open source. Users face fewer restrictions on what they can generate. Sacks argues that freedom explains its popularity, even if the models still carry Chinese state censorship on sensitive political topics.
The administration’s own policy swings have not helped calm the waters. In June the White House imposed temporary export controls on Anthropic models, then lifted them after negotiations. It also issued an executive order asking companies to submit new models for voluntary 30-day government review. Critics called the moves confusing at best. Alex Stamos, a security expert, described the flip-flopping as “about the dumbest thing they could possibly do” in comments reported by Politico.
Matt Pearl, a former National Security Council official focused on emerging technology, offered a more measured view. The administration faces an “extremely difficult and complex balancing act,” he said. Isaac Evans, another security researcher, warned that open-weight models let adversaries and criminals experiment freely. The net effect, several analysts argue, hands China extra time to close the remaining gap.
New Chinese releases keep coming. Moonshot’s Kimi follows recent updates from DeepSeek, Zhipu, ByteDance, and Ant Group. Many claim performance close to or matching leading U.S. systems at fractions of the cost. One analysis found DeepSeek’s latest model delivers comparable results to Anthropic’s Claude Opus 4.6 while charging 11 times less for input tokens and 28 times less for output. Such efficiency lets developers in emerging markets and cost-conscious U.S. startups bypass expensive domestic options.
Adoption numbers tell part of the story. Crypto executive Brian Armstrong and AI startup founder Flo Crivello have publicly praised Chinese models for cutting expenses. Cursor built part of its Composer 2 system on Kimi before its $60 billion acquisition by SpaceX. Airbnb says it relies mostly on U.S. models but uses a limited number of approved Chinese open-source versions through domestic providers. Government agencies have banned some Chinese tools, including DeepSeek. Private companies face no outright prohibition.
That gap creates its own headache. Procurement rules could discourage federal contractors from touching Chinese AI. Yet Kyle Chan, a fellow at the Brookings Institution’s China center, told CNBC it remains “ultimately impossible to ban China’s open-source AI models because their model weights are available freely on the internet.” Any broad restriction would run into First Amendment questions and risk hurting American startups that depend on open technology.
Daniel Remler, a senior fellow at the Center for a New American Security, sees the tension clearly. The Trump administration is “clearly worried” about the risks. Still, “restricting their use is going to be difficult.” He expects both the White House and Congress to keep warning companies away. Whether those warnings work is another matter.
New York’s recent ban on new data centers adds another layer of complication. Public distrust of big AI companies has grown, according to a Pew Research Center survey released in June. Many Americans, the data suggest, would show little sympathy if Washington stepped in to shield OpenAI or Anthropic from competition. That sentiment gives Sacks’s open-source stance unexpected popular resonance even as his formal influence inside the administration has waned.
The larger picture looks messy. U.S. policy veers between openness and control. Chinese labs release model after model that undercuts American pricing while matching or approaching performance. Advisers trade barbs on social media instead of presenting a united front. And the global market watches.
Beijing has signaled it may limit overseas access to its strongest models. That move could blunt some of the competitive pressure. But the open-weight versions already circulating will be hard to recall. Their weights sit on servers and torrent sites worldwide.
Trump faces a choice. He can double down on security reviews and procurement barriers. He can lean into the open-source advocates and try to match China’s speed and price. Or he can attempt some mix that satisfies neither side. The weekend’s public arguments suggest the internal debate has not produced a clear answer.
Meanwhile the models keep improving. Kimi is not the last surprise from Chinese labs. Each new release tests American assumptions about compute advantages, about the value of closed systems, and about how much government should steer commercial technology. The fractures within Trump’s AI world are not abstract. They shape what U.S. companies can build, what customers can afford, and who ultimately writes the rules for the next generation of artificial intelligence.
WebProNews is an iEntry Publication