A new consensus has emerged among some of the most important names in AI, and it is about the most alarming question in the field: could AI kill us all?
Anthropic CEO Dario Amodei announced that his company plans to slow its development of cutting-edge AI. Then Sam Altman of OpenAI and Elon Musk of xAI said they agreed. The change came after a researcher resigned from Anthropic, warning that top AI companies were advancing the technology at a dangerous speed.
The Free Press says the slowdown is worth taking seriously and interrogating thoroughly. Niall Ferguson will examine the latest unnerving developments, the forces driving the chances of catastrophe higher, and what, if anything, can be done about it. Tyler Cowen, by contrast, remains skeptical of the doomsaying and argues that the slowdown announced by the AI titans is not what it seems.
That split matters. When the leaders of the biggest AI companies more or less concede that speed itself is the problem, the debate stops being about distant speculation and becomes a question of governance. If the companies closest to the frontier are publicly signaling caution, then the market is already telling us something about the incentives at work.
The record also shows how fast the terms of debate can change. Last year, a Federal Reserve Bank of Dallas chart laid out a bewilderingly wide range of outcomes: never-before-seen prosperity, normal economic growth, or the end of the human race. That is not a normal policy spectrum. It is a warning that the upside and downside are both enormous, and that the institutions responsible for oversight are racing an invention they do not fully control.
There is a second thread here, and it is easy to miss. The same issue page places AI alongside energy abundance, American industry, and a conversation about tragedy and social unity. That is not an accident. AI is not merely a technical breakthrough; it is a force that will reach into power demand, industrial capacity, labor markets, and national security.
Follow the incentive, not the press release. If the biggest players are slowing down, it may be because the costs of unchecked acceleration are becoming harder to ignore. It may also be because public concern is rising faster than the companies can manage. Either way, the consequences are now large enough that no serious reader should accept the usual euphemisms about 'progress' without asking what, exactly, is being traded away.
What is true does not need an adjective. A technology that can produce prosperity on a historic scale can also produce catastrophe on a historic scale. The question is not whether the race is exciting. The question is who bears the risk when the race outruns prudence.
