Business leaders are starting to sound the alarm that American startups that supply datasets to OpenAI, Anthropic, and even the U.S. military are simultaneously selling the same refined data to Chinese AI labs. A recent Forbes investigation details how this $500 million annual data supply chain is helping Beijing close the gap in the AI race undercutting policy that restricts advanced chips but leaves data largely unregulated.
The same Silicon Valley firms that supply OpenAI, Anthropic, and federal entities including the U.S. Army and Air Force are also providing datasets to Tencent, Ant Group, Alibaba, ByteDance, and others. Top Chinese labs spend about $500 million annually purchasing the advanced instruction sets first developed for American companies that train AI to perform difficult tasks like coding, financial modeling, and cybersecurity. Chinese buyers treat themselves as a collective pool seeking the same top-tier Western datasets, allowing them to bypass costly trial-and-error.
Advertisement
Surge AI has supplied data to the Army, Air Force, and Chinese tech giant Tencent. Mercor holds a federal contract while also serving Tencent. AfterQuery earns recurring revenue from Chinese firms Ant Group and Alibaba, and Turing has worked with ByteDance (TikTok’s Chinese parent). Tencent was previously flagged by the U.S. government as linked to the Chinese military; their access to American innovation in cybersecurity and self-improving AI systems should be alarming to a Trump Administration hawkish on China.
This occurs against a backdrop where Artificial Intelligence and technology figures have gained substantial influence within the political right in recent years. In the Forbes report, Micro1 CEO Ali Ansari called the practice shameful, arguing that firms cannot credibly claim to support American AI dominance while selling millions in data to adversarial nations. Results of this transfer are already visible in Chinese models that rival or undercut U.S. performance at lower cost.
The focus of U.S. export controls thus far has been on semiconductors; data, however, remains largely unregulated within AI industries. High-quality training data is repeatedly identified as the highest-leverage input after computation itself. By packaging American expertise into resellable off-the-shelf datasets, these startups give Chinese labs a shortcut that chip restrictions alone cannot stop.
Advertisement
The Department of Commerce is the natural locus for addressing this gap through its Bureau of Industry and Security; the Department of Commerce administers the Export Administration Regulations governing dual-use technologies and emerging capabilities critical to national security. It already oversees licensing for advanced chips and has used authorities under the Export Control Reform Act to restrict AI model access in other contexts. Extending scrutiny to high-value training data exports would fit squarely within that mandate.
RedState reached out to Commerce Secretary Howard Lutnick and the Department’s Public Affairs office for comment on whether the administration is examining data sales of this kind and whether existing authorities could be applied more aggressively. As of publication, no response had been received.
The Trump administration has wisely prioritized preventing China from leveraging U.S. technology to compete against us. Restricting chips while permitting the free flow of the data those chips process leaves America with an incomplete safeguard. Closing that loophole would ensure that American innovation strengthens American capabilities rather than accelerating a rival’s in real time.
Advertisement
Editor’s Note: Do you enjoy RedState’s conservative reporting that takes on the radical Left and woke media? Support our work so that we can continue to bring you the truth.
Join RedState VIP and use promo code FIGHT to receive 60% off your membership.