There is no need to frame this as an “AI apocalypse.” The more serious issue is this: Researchers from inside the system are saying that the risks are real, while at the same time acknowledging that competitive pressure makes it very difficult for any single actor to stop on its own. And it is not only Coxon, who resigned, saying this. Evan Hubinger, who leads alignment research at Anthropic, has remained at the company while saying that Coxon is right on this point, that he personally puts the probability of AI causing human extinction within the next decade at above 10%, and that they still do not have a clear plan for safely aligning superintelligence. That, in my view, is where the real source of the problem lies. Throughout history, dangerous races have often emerged not from malicious actors, but from perfectly rational ones thinking: “If I stop, the other side will move ahead.” We should also avoid dismissing conspiracy theories categorically. But in this case, a conspiracy is not even necessary. Without any centralized plan, companies and governments can move in the same direction according to their own interests and eventually create a concentration of power that no one originally designed. So what happens if that power is badly governed? Let’s list some possibilities. Concentration of power: If highly advanced AI infrastructure becomes concentrated in the hands of a small number of companies and governments, information, capital, defense, media, and decision-making capacity could converge in the same centers. That would create a form of power far beyond a conventional monopoly. Surveillance and behavioral control: When AI is combined with facial recognition, location data, habits, purchasing behavior, private communications, and social network data, it can be used not only to observe people, but also to predict and influence their behavior. Political manipulation: Personalized propaganda, millions of different messages tailored to millions of different voters, synthetic video and audio, automated accounts, and psychological profiling could erode the shared reality on which democracy depends. Scaling cyberattacks: AI agents can greatly expand the ability to discover vulnerabilities, write code, conduct phishing campaigns, generate malware, and automate attacks. We do not even need to wait for some form of “superintelligent AI” for this. Autonomous weapons: If systems that are partially removed from the human decision loop begin selecting targets, the cost of warfare may fall for the attacking side. That could make war easier to initiate. Democratization of biological and chemical risk: Advanced models may lower the barrier to accessing harmful knowledge that previously required specialized expertise. The real danger is not necessarily that “AI builds a laboratory,” but that it makes dangerous knowledge easier to access and use. Jobs and income shock: If AI dramatically reduces the cost of many white-collar tasks, productivity may rise while wealth distribution worsens. The gains could accumulate in a handful of companies while the losses are spread across millions of workers. Erosion of human capability: If writing, thinking, research, decision-making, and memory are continually outsourced to AI, individual competence may weaken over time. That may be less an economic risk than a civilizational one. A deepening crisis of reality: If “Is this video real?”, “Is this document fake?”, and “Did this person actually say this?” become permanent questions, society may lose its common basis for verification. Dependency: As governments, companies, and individuals connect more critical decisions to AI systems, system errors can have much larger consequences. A small mistake in finance, health, energy, or defense can trigger cascading failures. And none of this requires superintelligence. A somewhat more powerful, more integrated, and more centralized version of today’s technologies may already be enough. So perhaps the real danger is not that AI itself will rule humanity. The real danger may be what the people who control this power choose to do with it. And at that point, the wealthy and powerful carry an enormous responsibility. But if they are still trapped in the same old obsessions — interest-based finance, debt, luxury homes and cars, endless accumulation, and the pursuit of ever more wealth — then humanity may have little choice but to pass through a period of serious disorder before it learns how to handle this kind of power.