Plato observed that those who tell the stories rule society.
Today, some of the world’s most powerful technology leaders are telling a remarkably consequential story about artificial intelligence—and increasingly, they are talking not about what AI can do, but about who will control intelligence, compute, labor, and ultimately national power.
I reviewed the recent public commentary of 20 leaders shaping the global AI ecosystem—from Mark Zuckerberg and Sam Altman to Jensen Huang, Satya Nadella, Alex Karp, Liang Wenfeng and Pony Ma. They disagree profoundly about where AI is taking us. But taken together, their pronouncements reveal an important shift: superintelligence is moving from speculation to planning assumption, and intelligence itself is becoming a strategic resource.
For the national-security community, that shift has another dimension. If intelligence becomes dramatically cheaper and more abundant, the benefits will not accrue only to the United States and its allies. They will accrue to our adversaries as well—and potentially disproportionately.
Here are eight observations.
- Superintelligence is becoming a planning assumption – leaders of the world’s largest AI investments are talking as if superintelligence is now a planning assumption. Mustafa Suleyman, CEO of Microsoft AI, says we have crossed an inflection point on the path toward superintelligence and argues that the important question is no longer simply whether it arrives, but what kind of superintelligence society chooses to build.
- The argument is shifting from capability to control — Mark Zuckerberg’s position is essentially a small number of superintelligences controlled by institutions is dangerous, but billions of individuals possessing superintelligent tools is safer and more prosperous. No surprise, but this is not universally accepted. Anthropic’s worldview emphasizes much more concern about proliferating dangerous capabilities leading to catastrophic biological, cyber, and geopolitical risks. Fairly clear that security-oriented leaders believe capabilities cannot safely proliferate.
- AI is becoming an instrument of national power — Jensen Huang’s “sovereign AI” argument is that nations need domestic capacity to transform their own data, language, and industrial base into intelligence. Nvidia describes data centers as “AI factories.” Mistral’s European positioning, DeepSeek’s Chinese open-model strategy, U.S. hyperscaler investment, and Gulf-state sovereign AI programs all point in the same direction. Cognitive capacity is becoming an element of national power. In the AI era, electricity, chips, data centers, models, talent, proprietary data, autonomous agents and robotics increasingly belong alongside traditional measures of strategic capacity.
- AI is changing the economics of human labor – a clean divide. Amodei warns that AI could eliminate roughly half of entry-level white-collar jobs and potentially push unemployment dramatically higher. Huang is the optimist. He argues that AI will ultimately produce more jobs, following the historical pattern of technological innovation. Andrew Jassy, Amazon CEO, sits in the middle — AI will create new work, but he told Amazon employees directly that the company expects AI efficiencies to reduce its corporate workforce over the next several years. Marc Benioff, Salesforce CEO, has perhaps supplied the most important vocabulary, stating that AI agents are “digital labor.” There is a different question related to national security. What happens to political stability when cognitive capital becomes dramatically cheaper than human labor?
- Agents move AI from answering to acting – we are rapidly moving from humans operating software to humans supervising AI agents that operate software to AI agents interacting autonomously with other AI agents. For national security, this is a critical threshold. The consequential transition may occur before AGI arrives: when autonomous systems can continuously conduct cyber operations, analyze intelligence, manipulate information, move money or control physical systems at speeds humans cannot match.
- Trust and authority become the scarce resource – Apple’s position is AI will know extraordinary amounts about you; therefore, trust and privacy become core architecture. Zuckerberg and Tim Cook, Apple CEO, both agree AI will become deeply personalized, but Meta believes we should democratize intelligence and give individuals agents, and Apple says put intelligence into a tightly controlled privacy architecture surrounding the individual. It is a debate over privacy and ultimately authority. AI eventually receives authority to communicate, spend, negotiate, execute software, and control machines.
The emerging AI contest isn’t simply America versus China, nor OpenAI versus Meta, nor even humans versus machines. It is becoming a contest over the architecture of power in a world where intelligence itself is an industrial resource. And that leads to the seventh area of interest.
- AI disproportionately empowers asymmetric adversaries – AI is a compression mechanism for state capacity. Countries and groups that historically lacked sufficient software engineers, intelligence analysts, linguists, scientists, propagandists or cyber operators can increasingly compensate with machines. That creates an uncomfortable possibility: the United States can lead the world in AI and still see the relative capabilities of its adversaries increase. The central policy question may be which capabilities can be widely distributed, which require safeguards, and which become dangerous when they cross certain thresholds.
- Intelligence services face a new counterintelligence problem – resource constraints drastically lower for adversaries. Massive datasets can be analyzed for behavioral anomalies. Social networks can be reconstructed automatically. Surveillance footage can be searched continuously. Open-source intelligence can be synthesized at machine speed. Synthetic identities can communicate convincingly with potential intelligence targets. Language barriers become less important.
Summary
The United States possesses extraordinary advantages—capital, semiconductor technology, universities, entrepreneurs, hyperscale cloud infrastructure, military alliances and many of the world’s leading AI companies. But authoritarian competitors possess different advantages.
For Western intelligence services, the concern is not simply that adversaries will conduct better analysis. The defining characteristic of the coming era may be that intelligence—technical, linguistic, analytical and eventually operational—stops being scarce.
That possibility changes the strategic equation. For the United States, abundant intelligence could accelerate scientific discovery, economic growth and military capability. But it could simultaneously give adversarial states and non-state actors access to capabilities previously constrained by money, expertise and organizational capacity.
The AI competition therefore cannot be measured simply by asking whether America has the world’s best model. Washington must ask who controls the compute, who possesses advanced models, who can distribute them, what authority autonomous agents receive, which capabilities proliferate to adversaries—and which societies prove most resilient to the disruption that follows.
The United States may ultimately win the race to build the world’s most powerful artificial intelligence yet lose the larger strategic contest. The real competition is over what happens when intelligence itself becomes abundant—and whether open societies or their adversaries are better positioned to exploit that abundance.
The race is on.
Submitted by Bob Pearson – Chair, Pearson Advisory Group and Member, Information Professionals Association