The Specific Event

Nvidia's board announced this week that it is opening a formal investigation into CEO Jensen Huang's relationship with Alice Huang, an executive at Megaspeed who was reportedly recruited by China's Ministry of State Security to conduct intelligence operations targeting Nvidia. Jensen Huang has separately confirmed that he was unfaithful to his wife and issued a public apology to his family. The board's statement signals that the investigation is treating this as both a personal conduct matter and a potential national security exposure. For a company that sits at the center of the global AI hardware supply chain, the implications extend well beyond executive misconduct.

I want to focus not on the personal dimension of this story, but on what the board's response reveals about a structural problem in how organizations govern their most consequential relationships: those between senior leadership and the external actors who seek access to organizational intelligence through interpersonal channels.

The Invisible Cage Problem at the Executive Level

Rahman's (2021) concept of the invisible cage describes how algorithmic control systems constrain worker behavior through opacity. Workers operate under constraints they cannot fully see, and that asymmetry systematically disadvantages them relative to the platform. What the Nvidia situation surfaces is a parallel dynamic at the organizational level, one that has received far less theoretical attention. National state actors, particularly intelligence services, construct influence relationships that are deliberately opaque to the target organization. The executive being cultivated is often the last person positioned to identify what is happening, precisely because the relationship is designed to feel legitimate.

This is not primarily a failure of Jensen Huang's judgment, though that is a separate question. It is a structural governance gap. Boards are designed to monitor financial performance, strategic direction, and legal compliance. They are not designed to detect the slow construction of influence relationships that never appear on a balance sheet or in a compliance filing. The organizational theory literature on executive oversight (see Kellogg, Valentine, and Christin, 2020, for related dynamics in algorithmic work contexts) has not seriously grappled with the fact that the most consequential information leakage in high-value organizations may travel through social relationships rather than through data systems.

Why Routine Governance Fails in Novel Threat Environments

The distinction Hatano and Inagaki (1986) draw between routine expertise and adaptive expertise applies here in a way that most corporate governance discussions miss. Boards develop routine expertise in the conditions they encounter repeatedly: financial fraud, regulatory violations, succession planning. They apply procedural knowledge because the problem structures are familiar. Foreign intelligence cultivation of senior executives is not a routine governance problem. It requires adaptive expertise, the ability to recognize a novel threat structure and reason from principles rather than precedent.

Nvidia's board is now doing something it almost certainly had no prior procedure for. The investigation it has opened is an improvised response to a situation that existing governance frameworks did not anticipate at this level of specificity. The National Security Agency and the FBI have frameworks for this. Corporate boards do not. That asymmetry is the governance gap the Huang situation has exposed.

The Folk Theory Problem in Executive Risk Assessment

One of the core distinctions in my own research concerns the difference between folk theories and structural schemas. Folk theories are informal, individually constructed models of how a system works. Structural schemas are accurate representations of the underlying causal architecture. Most executives, and most boards, hold folk theories about foreign intelligence operations. Those folk theories are built from popular media, general awareness, and occasional security briefings. They rarely reflect accurate structural understanding of how influence recruitment actually proceeds.

Gagrain, Naab, and Grub (2024) find that algorithmic media users who rely on folk theories of platform behavior show systematically worse outcomes than those who develop accurate structural schemas. The governance parallel is direct. Organizations whose boards rely on folk theories of geopolitical intelligence risk will consistently misidentify the threat, respond too late, and frame the problem in the wrong terms. Treating the Huang situation primarily as a personal conduct matter rather than as a systemic governance failure is precisely the kind of folk-theory response that leaves the underlying structural vulnerability intact.

What Organizational Theory Needs to Do Next

The Nvidia situation is not unique. It is the highest-profile recent instance of a category of risk that will only grow as AI hardware, AI model weights, and AI infrastructure become primary targets of state-level intelligence activity. Organizational theory has sophisticated frameworks for financial governance, algorithmic management, and platform coordination. It does not have an equivalent framework for governing the interpersonal channels through which organizational intelligence travels outside formal systems.

That is the problem worth taking seriously here. Not Jensen Huang's choices, but the board structure that had no mechanism to detect, monitor, or intervene in the kind of relationship that now requires a formal investigation. Building that framework is a genuinely open theoretical problem, and the events of this week make it harder to defer.