A growing number of researchers inside the frontier labs themselves, not just outside critics, are now voicing extinction-level concerns publicly. DeepMind research scientist Neel Nanda has said he believes there's at least a 10% chance AI could lead to human extinction, calling that figure “ridiculously high,” while OpenAI alignment research engineer Juan Felipe Ceron Uribe described frontier labs as “racing each other, kind of blindfolded.”
What makes this round of warnings different from earlier ones is the source: when researchers employed by the labs building the most capable models are the ones raising alarms, it's harder to dismiss the concern as coming from parties with no visibility into what's actually happening inside frontier development, which is precisely the criticism usually leveled at outside AI safety advocates.