US Charges Three Over Alleged AI Server Diversion to China in Export Control Enforcement

US authorities have unsealed an indictment against three men accused of conspiring to divert advanced artificial intelligence server technology from the United States to China in breach of export control laws. Announced on 19 March 2026, the case centres on allegations that high performance servers containing controlled AI graphics processing units were routed through Taiwan and Southeast Asia before reaching Chinese customers, with prosecutors claiming the operation was supported by false paperwork, concealed shipping arrangements and staged inspections.

The case is significant because it shows how aggressively US enforcement agencies are now treating the alleged diversion of advanced AI hardware. The Department of Justice, FBI and Department of Commerce’s Bureau of Industry and Security are framing the matter not as a technical compliance breach but as a national security threat. That reflects the strategic importance now attached to AI accelerator chips and servers incorporating them, particularly where regulators believe those products could enhance military capability, surveillance capacity or proliferation risk in China.

According to the indictment, the alleged scheme involved Yih-Shyan “Wally” Liaw, a US citizen and senior executive at a publicly traded US server manufacturer, Ruei-Tsang “Steven” Chang, a manager in the company’s Taiwan office, and Ting-Wei “Willy” Sun, described as a broker and fixer. Prosecutors allege the defendants worked with a Southeast Asian company and China-linked intermediaries to create the appearance of lawful sales to an approved end user, while the servers were in fact destined for customers in China. Chang remains a fugitive, while Liaw and Sun were arrested on 19 March 2026.

What makes the allegations especially striking is the scale. Prosecutors claim that between 2024 and 2025 the Southeast Asian company purchased around $2.5 billion of servers from the US manufacturer, many assembled in the United States, and that at least $510 million worth of servers were diverted to China in just a short period between late April 2025 and mid-May 2025. If proved, that would make this one of the more substantial alleged AI-related export diversion schemes publicly charged to date.

The indictment also underscores the increasingly sophisticated concealment methods that enforcement agencies say are being used to evade controls on advanced computing hardware. Prosecutors allege the conspirators staged thousands of dummy servers, used hair dryers to remove and reapply labels and serial number stickers, repackaged equipment in branded boxes and relied on encrypted messaging apps to coordinate shipments and internal cover stories. Those allegations will resonate across the technology supply chain because they suggest compliance risks now extend beyond paperwork failures into deliberate deception around audits, inspections and end-user verification.

For manufacturers, distributors and logistics providers, the practical lesson is that export control risk around AI infrastructure is no longer confined to semiconductor makers. Server assemblers, resellers, regional offices, brokers and warehousing partners are all in scope. The case reinforces the need for robust end-use and end-user diligence, closer scrutiny of transshipment routes through third countries, stronger controls over internal allocation approvals and careful validation of audit evidence where high-risk geographies or products are involved.

From a policy perspective, the charges illustrate how export controls have become a central tool in the wider US-China technology competition. Washington is signalling that restrictions on advanced AI compute are not symbolic. They will be enforced through criminal prosecutions, public messaging and coordinated action across justice, intelligence and trade authorities. The effect is likely to be greater pressure on global hardware supply chains to demonstrate not just formal compliance but credible control over where advanced AI systems ultimately end up.

It is important to note that the indictment contains allegations only, and all facts described by prosecutors remain unproven unless established in court. Still, the message to the market is already clear. Enforcement around advanced AI exports is intensifying, and companies exposed to high performance computing, semiconductors, cloud infrastructure and cross-border technology distribution should expect much closer scrutiny.

Read the press release here.