The moment when everyone begins talking about an issue, Ferguson writes, is usually the moment when the probability of rational decision-making goes to zero. His point is not subtle: AI is now everywhere, and the conversation around it may be drifting faster than the evidence.
Ferguson says the current mood has already produced a familiar response: pause, or at least 'pace the frontier.' He quotes Anthropic chief Dario Amodei saying, 'We must slow the pace at which we improve the capabilities of AI models.' He says Amodei warned that AI could find a cure for cancer, but also unleash cyberattacks and bioterrorism.
Ferguson also notes the strange speed of agreement. He says Sam Altman of OpenAI is postponing his company’s IPO until next year, that Elon Musk said 'Dario is right,' and that Demis Hassabis of DeepMind is on board too. In his telling, this is what alarm looks like when it moves from the fringe into the boardroom.
He then pushes the argument farther. The headline risk is not just 'the next 9/11, enabled by AI,' but also 'the next global financial crisis,' 'the next pandemic,' and even 'the next world war.' That is the scale of anxiety now attached to the technology.
Say it plainly: once fear becomes the organizing principle, institutions start rewarding caution over clarity. That is not the same thing as prudence. A system that slows itself because everyone is repeating the same worst-case scenario can become less innovative without becoming more secure.
The incentive problem is obvious. When the public conversation turns into a competition to sound most alarmed, the result is often more bureaucracy, more gatekeepers, and less serious judgment. The record is public on this point: panic rarely prices in trade-offs well.
That is why Ferguson’s warning matters even beyond AI. The issue is not only what the technology can do. It is what frightened people, regulators, and executives will do once they decide the future must be managed by committee. What is true does not need an adjective, and what is uncertain does not improve when the administrative state rushes to centralize control.
The free-market case here is simple. Innovation should not be throttled by reflexive fear, and democratic societies should be careful before they hand yet another frontier to a regime of caution, coordination, and permanent supervision. The right question is not whether AI can do harm. It can. The question is whether our institutions can still distinguish real risk from the current fashion of doom.
