Researchers Arvind Narayanan and Sayash Kapoor argue that AI safety debates overlook a third possibility beyond dismissing existential risk or treating it as settled fact: warnings about catastrophic AI risk may be sincere but mistaken and ultimately counterproductive. They contend society already underinvests in defenses against catastrophic and systemic risks, such as pandemics and cyberattacks, which AI can amplify regardless of whether existential-risk predictions prove correct. They warn that framing AI primarily around existential risk drives public concern but also risks partisan polarization and diverts resources from practical defenses. They call for a broader, “big-tent” safety movement that includes people who do not share existential-risk concerns but support measures like transparency and liability, rather than a narrower approach that could provoke backlash.
