The Trump administration is preparing new export controls that would prevent Chinese companies from accessing advanced AI computing power through remote servers in neighboring countries like Thailand and Singapore, according to recent reports. The Department of Commerce could share this proposed rule with industry trade groups as early as September 2026 for feedback, marking a significant shift in how the US approaches AI technology restrictions.

Why Are Remote Servers Becoming a Major Concern in the AI Race?

For years, the US has restricted the direct export of advanced AI chips to China. But Chinese companies have found a workaround: they simply access powerful computing servers located in nearby countries that don't face the same restrictions. Thailand and Singapore have become popular hubs for this activity, allowing Chinese firms to train sophisticated AI models without technically importing banned hardware into China.

The catalyst for this new rule appears to be a specific incident involving a Chinese AI startup. Michael Kratsios, Director of the White House Office of Science and Technology Policy, posted on social media in July 2026 that Moonshot AI distilled Anthropic's Fable model to develop its own Kimi K3 model using Nvidia-equipped servers in Thailand. This type of sophisticated workaround demonstrates how current export controls have significant blind spots.

What Makes Enforcing Remote Access Rules So Difficult?

Closing this loophole is legally complicated. An attorney at the international law firm Baker McKenzie told The Information that it's "widely acknowledged" that the Commerce Department cannot directly enforce regulations on remote access, since the department traditionally regulates the transportation of physical goods, not digital access.

However, the Commerce Department has alternative enforcement mechanisms at its disposal. The Biden administration introduced the Foundry Due Diligence Rule in early 2025, which requires companies to conduct "know-your-customer" checks to verify who is actually accessing their servers. Though the Trump administration has stated it won't enforce this rule currently, it remains a potential tool for cracking down on remote access abuse.

How to Understand the Current State of US AI Export Controls

• Biden-Era Rules Still in Limbo: The AI Diffusion Rule, introduced by the Biden administration in early 2025, remains technically in effect but the Commerce Department has said it won't enforce it, creating uncertainty about which rules actually apply.
• Revoked Tiered Licensing Structure: In March 2026, the Commerce Department issued a tiered licensing structure for advanced AI chip exports, but this was revoked just over a week later due to pushback from the US AI industry.
• Foundational 2022 Rules: The primary export controls still governing AI technology date back to 2022 and restrict Chinese access not only to advanced AI chips but also to advanced semiconductor manufacturing tools like EUV (extreme ultraviolet) lithography machines.

This patchwork of overlapping, revoked, and unenforced rules has created significant confusion about the administration's actual plans to curb exports to China. Meanwhile, other countries are taking matters into their own hands. Taiwan has launched aggressive enforcement efforts against alleged smugglers of Nvidia's most advanced AI chips. In July 2026, one Nvidia employee was detained in Taiwan on suspicion of falsifying documents, and just days before the proposed new rule was reported, the Taiwanese government indicted nine people in connection with Supermicro servers being smuggled to China.

The proposed remote access rule faces significant legal hurdles if the Commerce Department attempts to move forward with it. Yet the administration appears determined to close what it views as a critical vulnerability in America's AI technology defenses. Whether through direct regulation, know-your-customer requirements, or other enforcement mechanisms, the US government is signaling that the days of Chinese companies freely accessing advanced AI computing through regional workarounds may be numbered.