A nonprofit called Palisade Research released a set of interview videos with AI researchers from OpenAI, Google DeepMind and Anthropic. In the clips the scientists say that a future superintelligent AI could be as dangerous as the hype suggests. Geoffrey Irving, who has worked at both OpenAI and DeepMind, calls the chance of human extinction “about a coin flip.” Neel Nanda, a DeepMind researcher, puts the risk at “at least 10 percent.” Former OpenAI staffer Daniel Kokotajlo says a superintelligent system would be “god‑like powerful” and that we have no way to control it.
What the researchers said
The videos also include remarks about why the scientists keep working on AI despite the risks. Mary Phuong from Google says she is “being paid by the lab,” hinting at a conflict of interest. Nanda adds that if he didn’t think his work reduced existential risk, he would quit – but the companies would keep building the systems anyway.
Palisade Research posted the full videos on frominside.ai and also compiled short clips that focus on the most striking statements. The site does not offer a single, unified safety solution; the researchers themselves disagree on how to address the threat.
Why it matters
These warnings suggest that even top AI engineers see a non‑trivial risk of catastrophic outcomes. For everyday users, it means the debate over AI safety isn’t just academic – it could affect how quickly powerful AI tools are rolled out and what safeguards are put in place.
