Nvidia, Export Controls and the Corruption Risk Finance Desks Missed

On December 8, 2025, President Donald Trump gave Nvidia permission to sell its H200 chip to China — on condition that the U.S. government received a cut of the sales. That single arrangement, according to a Fortune commentary by Susan Ariel Aaronson, converted American export policy from a national security instrument into a revenue stream, and it sits at the center of a corruption case study that chief financial officers should treat as a governance warning, not a political footnote.

What did the 2025 UN panel actually predict?

In 2025, the UN Office of Drugs and Crime and the Organization of American States co-convened a panel to examine whether artificial intelligence constitutes the biggest corruption risk in history. The panelists feared AI would enable graft — easier money laundering, simpler bribes, faster disinformation. Events proved them half right. By November 2026, Aaronson argues, the corruption of AI had arrived, but it began not inside the technology. It began with its stewards: heads of state, government officials and technology chief executives steering AI governance itself.

The author applies Transparency International's definition — the abuse of entrusted authority for private gain — and names one company as the poster child for that abuse in the United States: Nvidia.

How did Nvidia obtain its China exemption?

Last year, the Trump administration restricted chip sales to China, citing fears that the chips would advantage the Chinese military or erode U.S. leadership in AI. Nvidia's then-chief executive, Jensen Huang, used flattery and donations to influence the administration's export policies, Aaronson writes. In December, the president granted Nvidia the H200 exemption, with the U.S. government taking a share of the proceeds. AMD also received an exemption.

The Constitutional problem is immediate. The Congressional Research Service reported that Article I, Section 9, Clause 5 of the U.S. Constitution prohibits federal export taxes and duties, and that 50 U.S.C. §4815(c) bars the Bureau of Industry and Security from collecting fees for considering or issuing export licenses. The arrangement, in other words, gave Washington a direct financial incentive to loosen controls on AI infrastructure to China — and did so through a mechanism two separate legal authorities appear to forbid.

What did officials know, and when?

The record does not stop at the White House door. Nvidia's general counsel, Tim Teter, appeared to interfere in an investigation of the company's sales to Chinese and Malaysian entities, according to a November 9, 2026 New York Times report the author cites. Commerce Secretary Howard Lutnick was warned about the dangers surrounding Nvidia and Huang, a Times report of November 12, 2026 established, yet took no action. Aaronson's verdict is blunt: this corruption and inaction put U.S. national security and the American people at risk.

Note what is absent. No enforcement action is recorded. No resignation is cited. The warnings produced nothing.

Why are export controls structurally vulnerable?

An OECD analysis Aaronson cites identifies the mechanism. Export controls sit at the intersection of high-value restricted goods, bureaucratic gatekeeping across customs, business and trade agencies, and complex cross-border supply chains. Officials issue licenses, classify goods and clear shipments — concentrated discretionary power that lets corrupt actors solicit bribes for faster approvals or a blind eye. U.S. agencies, meanwhile, struggle to coordinate trade, customs and anti-corruption efforts split across separate mandates.

Controls also manufacture scarcity. For dual-use items such as AI systems, some buyers will pay extreme prices for limited supply. That scarcity has produced an unusual coalition: Google, OpenAI, Amazon, Anthropic and Microsoft — three of them Nvidia's largest clients — support legislation that would further restrict Nvidia's chip exports to China, according to Politico and The Wall Street Journal reporting from November 12, 2026.

A Transparency International UK assessment adds a third weakness: opaque national security exemptions. Licensing tied to classified intelligence resists scrutiny by independent watchdogs. And the watchdogs themselves have thinned. In 2025, the Trump administration fired numerous inspectors general, including at the Department of Commerce, along with anti-corruption prosecutors.

Where does policy go from here?

Aaronson concludes that export controls are not the best tool to keep advanced AI from adversaries, precisely because they multiply corruption opportunities. Studies she cites show such controls rarely work, and Politico has reported that members of Congress find it nearly impossible to support loosening them — a policy trap with no exit through the normal channel. Current implementation, she adds, citing analyses from CSIS and CNAS, can damage U.S. economic and technological competitiveness.

Her alternative is modest: Congress should weigh a different approach, and the United States should coordinate with Western allies to manage technology transfers that threaten shared security interests. She concedes there are no easy answers.

For finance leaders, the takeaway is narrow and unsentimental: when a government monetizes its own restrictions, every company operating under those restrictions inherits a compliance risk that no policy manual currently prices.