According to The Decoder's report, in a survey of more than 1 500 leading AI researchers, the average estimated probability of an AI-related extinction scenario in 2024 was 18%.
The topic sparked renewed discussion following comments from Anthropic researcher Jacob Coxon and OpenAI researcher Daniel Selsam, as well as a former DeepMind researcher. In The Decoder's recounting, these remarks are presented as warnings about existential AI risks.
The news significance lies not in proving an inevitable catastrophe, but in the scale of disagreements and concerns within the professional community. However, the source is available only as a brief meta-description, so survey details, participant composition, and the basis for claims of further increases in the estimate remain unverified.
editorial commentary
Why it matters
A likely consequence is increased demands for AI risk assessment and control, but available data does not directly confirm this. The next observable signals will be the publication of the survey methodology and new comparable estimates. The main uncertainties concern sample quality, the interpretation of the term 'extinction', and unconfirmed growth in the metric.