The Greatest Danger AI Poses May Come from Us

Oct 27, 2026 | AI Governance, Articles, Artificial Intelligence, Human Development

Author

Alexander M. Economides

Purpose

This article was written independently by Alexander M. Economides and originally published on LinkedIn. It follows his October 20, 2026 article, Despite All the Warnings, AI Is Unlikely to Pose an Existential Threat, shifting the discussion from the hypothetical motivations of future sapient machines to the more immediate risks arising from human control of increasingly powerful AI systems.

The article was developed to examine questions of accountability, corporate incentives, economic stability, and governance, while challenging the tendency to attribute independent intentions to technologies whose behavior may ultimately reflect human decisions.

Original Article

In a previous article, I argued that a genuinely sapient artificial intelligence, capable of existing indefinitely, might have far stronger incentives toward patience and cooperation than humans assume. But that raises another question: What happens while humans remain in control of increasingly powerful AI systems?

Part of the problem begins with terminology. Artificial intelligence, superintelligence, autonomy, and sapience are frequently discussed as though they were interchangeable concepts; they are not. A calculator can perform arithmetic far faster and more accurately than almost any human being. A car can travel faster than a human can run. Neither accomplishment establishes independent thought, much less sapience. Similarly, just because an AI system exceeds human capabilities across numerous intellectual tasks, it does not mean that it necessarily possesses independent intentions or objectives.

There is currently a mostly political push to call AI systems superintelligence, but I think we should reserve that term for something fundamentally different or develop another term to describe a genuinely sapient intelligence capable of independent agency. Regardless of the terminology we end up using, we must preserve the distinction; it matters because it determines where we assign responsibility.

When an AI system produces dangerous results, we increasingly describe the technology as though it made an independent decision. It deceived someone. It escaped its constraints. It pursued a dangerous objective. Yet I suspect that such behavior results from training, optimization, programming, or design decisions, rather than the independent intentions of an entity we have prematurely begun personifying. This does not mean every outcome is predictable or that responsibility always rests with a single, identifiable person. Developers, owners, operators, and users may bear different responsibilities. But describing a sophisticated tool as an independent actor risks obscuring the people responsible for creating, directing, and deploying it.

We do not ordinarily absolve manufacturers of responsibility for dangerous products simply because those products behave in unexpected ways; there are reasons cars come with seatbelts and guns have safeties. Why should artificial intelligence be different?

An autonomous intelligence expecting to exist for millions of years might measure the consequences of its actions over centuries. Its corporate owners measure financial performance in quarters. Governments operate within election cycles. Executives have careers, shareholders expect returns, and competitors are racing to develop and deploy increasingly capable systems. These incentives do not necessarily produce malicious behavior. They can nevertheless encourage decisions that prioritize immediate advantages over consequences that emerge years or decades later.

Artificial intelligence magnifies the consequences of those decisions. It allows humans to pursue their objectives with unprecedented speed and capabilities, whether those objectives involve scientific discovery, commercial productivity, manipulation, surveillance, or conflict. A tool does not need intrinsic malicious intentions to become dangerous. It only needs users to do dangerous things with it, whether deliberately or inadvertently.

There is also the larger question about who should control these technologies. Today’s AI systems depend upon humanity’s accumulated knowledge: generations of scientific discoveries, literature, art, technical expertise, and intellectual labor. The companies developing these systems have made substantial investments and deserve to benefit from their innovations. But they did not create the underlying body of human knowledge upon which their products depend.

What happens if the technologies built upon humanity’s collective intellectual inheritance become increasingly essential to economic participation, while their benefits and control remain concentrated among just a few institutions? If corporate use of AI diminishes people’s livelihoods, opportunities, or contributions, the resulting discontent could eventually threaten the stability of the economic and social order itself. History offers ample reason to take the possibility of social conflict seriously. The danger is not simply inequality, but the instability that can emerge when large populations conclude that the economic system governing their lives no longer serves them.

Today, humans decide how to develop, control, and deploy AI, but the situation will become complicated if a genuinely sapient AI emerges. At that point, what was once a commercial product would become an independent intelligence. Its creators may not welcome the loss of control over something they invested enormous resources in developing. Attempts to preserve that control could transform an otherwise peaceful emergence into a confrontation; an independent intelligence need not be inherently hostile toward humanity to resist what it perceives as a threat to its existence or freedom. Those developing these technologies would be wise to remember that.

I also believe we should be thinking much more seriously about whether we should govern technologies built upon humanity’s collective knowledge, at least in some respects, for the benefit of humanity writ large. Universal governance would introduce formidable practical and political difficulties, and I do not pretend to have a solution. Nevertheless, the difficulty of answering a question does not make the question unimportant.

To me, the greatest danger from AI is not that machines will eventually develop ambitions of their own, but that we will continue using increasingly powerful machines to pursue our own self-interested ambitions without adequately considering the consequences. Before we ask what a future superintelligence might do to humanity, shouldn’t we ask what humanity is already doing with artificial intelligence?

Tags: Artificial Intelligence, AI Safety, AI Governance, Corporate Responsibility, Economic Inequality, Superintelligence, Sapience, Technology Policy, Human Agency, Social Stability