Nvidia, AI Export Controls, and the Unspoken Duty of Tech Leaders

In an April 2024 interview with tech podcaster Dwarkesh Patel, Nvidia CEO Jensen Huang faced pointed questions about the national security implications of exporting advanced AI hardware to strategic adversaries. Huang’s reluctance to engage directly with these concerns has reignited a critical debate: What responsibility do American tech companies bear in ensuring their products do not undermine U.S. National security? The answer may lie in a long-dormant mechanism from the Cold War—a “transmission function” that once bridged industry and government on sensitive technology.

The Transmission Function: A Cold War Legacy

The concept of the “transmission function” traces back to the 1970s, when industry leaders like Texas Instruments’ J. Fred Bucy helped shape U.S. Export control policies. Bucy’s 1976 report, which reframed export controls around “militarily critical technologies” rather than end-products, laid the groundwork for the Export Administration Act of 1979. This framework prioritized controlling technologies over finished goods, a principle now central to U.S. Chip export restrictions on China.

Historically, this function relied on industry executives voluntarily informing the government about their products’ potential military applications. However, the Toshiba scandal of the 1980s—where Japanese and Norwegian firms sold advanced milling machines to the Soviet Union—highlighted the consequences of its absence. The U.S. Navy reportedly incurred up to $30 billion in costs to regain its acoustic advantage, underscoring the stakes of such oversight.

AI’s New Frontier: Anthropic, OpenAI, and the Limits of Self-Regulation

Today, the transmission function faces a new test in the AI sector. Companies like Anthropic and OpenAI have taken starkly different approaches to balancing commercial interests with national security concerns. Anthropic’s refusal to grant the Pentagon unrestricted access to its Claude model for autonomous weapons or mass surveillance has led to legal battles, with the company filing a lawsuit against a Pentagon designation of “supply chain risk.”

OpenAI, by contrast, has negotiated contractual safeguards that align with U.S. Laws like the Fourth Amendment, despite criticism that these protections may be weaker than they appear. Both companies acknowledge the risks of their technology, embodying the transmission function’s core principle: recognizing downstream security implications and addressing them proactively.

Yet Nvidia, the semiconductor giant underpinning much of the AI industry, has drawn sharp criticism for its stance. Huang’s dismissal of Patel’s questions as “childish” and his assertion that China already possesses sufficient compute capacity to develop advanced AI models has raised concerns about the company’s willingness to engage in this dialogue.

The Mythos Revelation and the Urgency of Coordination

Anthropic’s recent disclosure of the Claude Mythos Preview model—capable of identifying thousands of security vulnerabilities across major operating systems—has added urgency to the debate. The company’s Project Glasswing initiative, which aims to coordinate industry-wide fixes, highlights the need for collective action. However, Huang’s refusal to address the risks of frontier compute falling into adversarial hands suggests a disconnect between industry and government priorities.

Experts argue that the U.S. Must act swiftly. A 2024 report by the Center for Strategic and International Studies (CSIS) warns that Chinese AI models trained on domestic hardware could become the “default frontier substrate” in the global south, eroding American technological leadership. This scenario underscores the importance of a coordinated policy framework.

Rebuilding the Transmission Function: Lessons from the Past

The U.S. Government has taken steps to revive the transmission function. In January 2026, the Department of Defense merged the Defense Science Board and Defense Innovation Board into the Science, Technology, and Innovation Board, a move aimed at fostering closer collaboration between industry and policymakers. However, recent delays in implementing federal pre-deployment reviews for frontier AI models reveal ongoing challenges.

Historical precedents suggest that formal task forces, akin to the Bucy report, could provide a solution. Such a mechanism would require industry leaders to publicly disclose the