OpenAI and DeepMind staff say AI could kill everyone

A nonprofit called Palisade Research published frominside.ai on September 29: 12 filmed interviews with people who build or built frontier AI, in which several of them describe human extinction as a live possibility rather than a marketing line. Reuters was given the videos ahead of publication and reported the project that morning, writing that the current and former employees of OpenAI and Google DeepMind described their concerns about existential risk as sincere.

The interviews run from 10 minutes to 43, and the site sets the sharpest lines over each subject's face with a timestamp into the full cut.

Neel Nanda, a research scientist at Google DeepMind, puts the chance in a number: "I would put at least a ten percent chance that it causes human extinction, and that is ridiculously high." Geoffrey Irving, co-founder and chief scientist at the nonprofit Resolution, who has worked at both OpenAI and DeepMind, says: "The chance of human extinction is about a coin flip, in my view, about a half." He also told Reuters that the labs "are overplaying the extent to which this is a pure coordination problem. They could just stop unilaterally."

On whether the warnings are a marketing posture, the project gives two answers. Mary Phuong, a DeepMind researcher, says: "I think you absolutely should be suspicious of what I'm saying because I am being paid by the lab." Irving's reply to the same question is that "they wouldn't be saying, 'My product might kill everyone,' if they didn't actually believe that."

The interviews also cover the labs from the inside. Rosie Campbell, a policy researcher at OpenAI until 2024, told Reuters the organisation had grown more siloed during her tenure, which she said made its direction harder to influence. Vishal Maini, a DeepMind researcher, says that when he arrived, "external communication about the possibility of human extinction wasn't permitted by anyone at any level of the organization."

Palisade is not a neutral party and does not present itself as one. The group's FAQ quotes its executive director saying in his video, "The future is not yet decided. We have choices to make," and states that Palisade generally agrees the AI companies are "gambling with our lives."

What makes the project usable as source material is the production, not only the argument. The footage was shot in a rented San Francisco studio and in London, where a separate crew filmed interviews that the project lead conducted over Zoom, and the team says 15 days passed between the idea and publication. Every video is released under a Creative Commons Attribution 4.0 licence, with an invitation on the page to clip them.

The project lands in a week when the industry's own safety language is already in the news. Reuters reported on September 28 that Anthropic plans to warn potential investors in its initial public offering that advanced AI could pose "catastrophic or existential risks to humanity," and OpenAI said it had held back the release of a more powerful model.

Watch the collection at frominside.ai, or the trimmed cut on the Palisade Research YouTube channel. The full interviews are free to clip and reuse with credit.