Palisade Research Video Series Warns Superintelligence Risk Is 'Exactly as Dangerous as It Sounds'

A nonprofit called Palisade Research has released a video series in which roughly a dozen AI researchers warn that superintelligent systems carry risks serious enough to threaten human survival. The interviews include current and former employees of OpenAI, Google and Anthropic, according to The Verge, and the project's framing is blunt: the danger, in the group's telling, is exactly as dangerous as it sounds. The videos land as labs race to build more capable systems and as public scrutiny of their safety practices grows.

The most quoted line comes from Geoffrey Irving, who has worked at both OpenAI and Google DeepMind. In his interview he puts the odds of human extinction at roughly a coin flip. That is a far more pessimistic estimate than the cautious language typically used by lab executives, who tend to describe catastrophic risk as low-probability but worth mitigating. Irving's framing shifts the debate from abstract precaution to a claim about near-term odds.

Palisade Research describes itself as a nonprofit that studies how AI systems behave. The organization has drawn attention before for work on models that resist shutdown or otherwise act against operator intent, a line of inquiry that sits at the edge of mainstream safety research. By collecting on-camera testimony from people who have built frontier systems, the group is trying to move the argument out of technical papers and into a format that travels further.

The roster matters as much as the message. Voices from OpenAI, Google and Anthropic carry weight because those companies are training the largest models and setting the norms competitors follow. Former staff in particular can speak more freely than employees still bound by confidentiality and communications policies. Their participation also highlights a recurring tension inside the field: many of the people closest to the technology are among the most alarmed about where it leads.

The timing is not accidental. Governments are drafting AI rules, labs are publishing safety frameworks, and debates over evaluation and oversight have become routine. A video series aimed at a general audience competes for attention with product launches and funding news, which usually dominate coverage of the sector. Whether testimony from insiders changes public opinion or policy is an open question, but it raises the cost of dismissing existential risk as a fringe concern.

What to watch next is whether the companies named respond directly to the claims, and whether more current employees follow the former staffers onto camera. Palisade Research has not said how many more interviews it plans to release or how it selected participants. The broader test is whether stark warnings from insiders translate into concrete changes in how frontier models are trained, tested and deployed.