Narayanan and Kapoor advocate for a ‘big-tent’ approach to AI safety rather than one focused on the risk of extinction

The authors of *AI as Normal Technology* argue that the existential framing, which has become dominant, risks polarising the debate and overshadowing practical defences against risks that are already very real.

In an essay published on October 1, 2026 on their newsletter Normal Tech, Princeton researchers Arvind Narayanan and Sayash Kapoor take a stand in the debate in the AI security community. They reject both the idea that existential risk is imminent and the idea that alerts are manipulative: according to them, these warnings can be sincere while being erroneous and counterproductive.

Very real risks, but of a different nature

The two authors acknowledge that society underinvests in disaster protection, as shown by unpreparedness for pandemics. They see AI as an amplifier of existing systemic risks, especially in cybersecurity, where no culture of modeling extreme scenarios exists, or in the gradual erosion of institutions such as the press. They also point out that, in Gallup polls, barely half a percent of respondents cite AI as the most important problem.

Their fear: that the existential framework calls for only one response, prohibition, to the detriment of a hundred more modest measures of defense and resilience. They propose a "big tent" movement, pluralistic in its convictions as well as in its values, which would rely first and foremost on common ground such as transparency and legal accountability, rather than on a policy dictated by panic.