Do Most AI Researchers Actually Believe AI Poses an Existential Threat?

No Single, Unified Expert Consensus

Contrary to how the topic is sometimes portrayed in public discussion, AI researchers do not hold one unified view on whether advanced AI poses an existential threat to humanity. This is a genuinely contested question within the field, with credible, accomplished researchers occupying positions across a wide spectrum, from serious concern to considerable skepticism. Attempts to survey AI researchers’ views on this topic have generally found a spread of opinion rather than overwhelming agreement in either direction, reflecting the fact that existential risk from advanced AI involves significant uncertainty about future technological capabilities that don’t yet exist.

This divide is worth understanding clearly, since public discussion sometimes flattens the range of expert opinion into a simpler narrative than the actual state of the field reflects.

Voices of Serious Concern

Some prominent AI researchers, including individuals who have played significant roles in founding or leading major AI research organizations, have publicly expressed serious concern about long-term risks from advanced AI systems, and some have signed public statements calling for the topic to be treated as a global priority alongside other major societal risks. These researchers generally argue that the pace of AI capability improvement, combined with unresolved technical challenges around reliably aligning highly capable systems with human values, justifies significant caution and dedicated safety research.

Notably, this concern has come from within the AI research community itself, including from people actively involved in building increasingly capable AI systems, which some observers see as lending particular weight to these concerns, while others see it as an inherent tension worth scrutinizing.

Voices of Skepticism

Other equally credible researchers are skeptical of existential risk framing, for a range of reasons. Some argue that the specific scenarios described are too speculative or poorly specified to usefully guide policy or research priorities today. Others argue that focusing heavily on long-term, hypothetical catastrophic risk can distract attention and resources away from more immediate, already-documented AI harms, such as bias, misinformation, labor market disruption, or concentration of power among a small number of AI companies. This skepticism generally isn’t a claim that AI poses no risks at all, but rather a disagreement about the plausibility, timeline, or priority of the specific existential risk framing.

Bottom Line

There is no clear consensus among AI researchers that AI poses an existential threat; expert opinion is genuinely and substantively divided, with credible researchers on both sides of the debate. Surveys of the field generally reflect this spread of views rather than uniform agreement, and understanding this divide is important for interpreting media coverage or public claims about what “AI researchers believe.”