Jensen Huang’s resignation from Nvidia on November 12, 2026, wiped hundreds of billions of dollars from the company’s value and exposed how much the AI industry depends on a few individuals. Control Risks advises treating critical-person dependencies, at your own firm and at key suppliers, with the same rigour as technology and infrastructure.
Why is critical-person risk missing from AI concentration assessments?
Organisations today rightly assess concentration risks in terms of chips, compute, cloud and models, but they too often overlook people. Critical-person risk is when technical knowledge, commercial relationships and decision-making authority become heavily concentrated around particular individuals, impacting national competitiveness, strategic capability and corporate resilience. The Securities and Exchange Commission has long deemed leadership succession in public companies as a core Board risk mitigation duty, reflecting the need to avoid strategic disruption from a sudden leadership vacancy. Yet even businesses that plan for the loss of their own critical people may overlook equivalent dependencies at their technology suppliers.
Nvidia is a clear example of critical-person dependency. A founder-CEO led the company for more than three decades, ran a flat structure with dozens of direct reports and had no publicly disclosed succession plan – indeed, he was famously against such planning. Leadership change does not automatically mean operational disruption, and Nvidia’s board expects its roadmap and customer commitments to continue. Apple showed in 2011, when Tim Cook succeeded Steve Jobs, that even the rock-star founder-CEO is replaceable. Even so, given the centrality of Nvidia in so many corporate sectors, its leadership crisis affects more than just Nvidia.
Where else does critical-person dependency sit in the AI industry?
Critical-person dependency in AI goes well beyond any one company’s leadership, and it does not stop at the company’s perimeter. Engineers and researchers at chipmakers, AI labs and cloud providers hold scarce technical knowledge that the rest of the industry relies on. Finance and commercial leaders shape investment and partnership decisions with consequences across the technology supply chain. Relationships with key customers, suppliers and governments often rest with a few people rather than with the institution. Because so much of the AI stack runs through a small number of firms, losing or compromising a handful of individuals in one of them can have consequences for the entire ecosystem.
The qualities that make key individuals in AI critical to their organisations also make them attractive to a range of actors. Huang’s departure followed reports of an alleged foreign intelligence approach. Those allegations have been denied, and Nvidia’s board said it had no reason to believe any sensitive information had been compromised. But the episode is a reminder that concentrated knowledge and access create both continuity risk and opportunities for adversaries.
Critical-person exposure is not confined to the most senior people in businesses. Competitors, state and criminal actors seek access to sensitive assets throughout an organisation. A research engineer with access to model weights, a cloud administrator with privileged credentials or a commercial lead who knows which customers are negotiating what can be more useful to a threat actor than a chief executive. And a susceptible person need not have specialised engineering skills or senior standing: the CEO’s executive assistant, or the after-hours cleaning crew, are common targets for foreign intelligence services. Intelligence services may seek technology and strategic insight to advance national programmes; competitors seek proprietary and commercial advantage; criminal groups seek data, leverage and money. Their objectives differ, but each has reason to identify and cultivate people with valuable access.
How do adversaries approach critical people?
Adversaries’ methods range from the technical to the personal. Cyber attacks and social engineering remain scalable routes into organisations, increasingly supported by AI-enabled automation. But money, ideology, coercion, ego and of course passion, remain effective levers to manipulate vulnerable people toward ulterior motives. As technical defences improve and automation overtakes digital spaces, Control Risks expects to see a parallel growth in human approaches as resources free up to undertake these. These include recruitment pitches presented as job offers or advisory roles, relationships cultivated patiently over months or years through professional and personal networks, and, where those fail, bribery or coercion. Travel adds further risk, because devices, accommodation and local contacts are harder to control abroad.
Most approaches to critical people arrive through a wider ecosystem rather than directly, a customer seeking more technical detail than a contract requires, a supplier or contractor with privileged access and weaker security, an investor or recruiter whose backers are unclear, or a new contact on a professional networking site. Each gives an adversary a way to build familiarity before anything sensitive is requested. An approach may begin as a legitimate-looking collaboration, with requests gradually moving towards unpublished research, sensitive infrastructure or confidential commercial information. The warning often lies in the pattern: growing demands for access, unusually generous rewards and increasing requests for discretion.
The systemic consequences of a shock at a critical part of the technology ecosystem like Nvidia are hard to assess today and the company has been clear that it does not anticipate a fundamental shift in its current business. Even so, the resignation should serve as a warning bell. The current AI ecosystem is highly dependent on the continuity of supply of many single-vendor products. Worse still, multi-year compute commitments which power much of the industry’s current capital allocation and future promises are highly contingent on a set of economic and political realities which may rapidly change.
A five-question framework for assessing critical-person risk across AI suppliers
Inside their own walls, organisations should apply the same discipline. Better succession planning and more executive protection are both part of the answer, but neither is enough on its own. Control Risks’ Critical-Person Risk Framework asks five questions:
- Who is critical, including beyond senior leadership?
- What knowledge, access, relationships and authority sit with them?
- How the organisation would continue if they were unavailable, whether through departure, incapacity or compromise?
- Which actors could benefit from accessing, influencing or compromising them, and how they are likely to go about it?
- Where exposure exists across their digital, physical, professional and personal environments?
This framework should be applied to every systemically important vendor and supplier in a company’s ecosystem.
How should organisations and boards respond?
Resilience and protection should be designed together. On the resilience side, that means spreading essential knowledge, documenting key relationships and testing continuity arrangements before they are needed. Protection is harder to do well, because it has to draw on intelligence about likely adversaries to coordinate cyber, physical and personnel security, and to prepare the people most likely to be approached. Boards should know how they would respond, and how quickly, if a critical individual were compromised or had to step away without warning, whether that individual is their own or sits at a supplier they depend on.
Organisations can begin with three steps:
- Map the dependencies. List the critical people in your own business and at your five most systemically important technology suppliers, using questions 1 and 2 of the framework.
- Test the scenarios. Run the remaining questions against your three most important suppliers, then rehearse what happens if a critical individual departs, is incapacitated or is compromised without warning.
- Take it to the board. Add critical-person risk to the risk register and report the findings to the board or risk committee.
Control Risks helps organisations assess critical-person risk across their technology supply chains. To discuss how the Critical-Person Risk Framework applies to your organisation, contact our team.