UN AI experts warn against 'apocalyptic' rhetoric on risks

Several United Nations AI committee experts cautioned against extreme language used by some sector professionals regarding artificial intelligence risks. Executive Joelle Barral of Google DeepMind, a member of the International Scientific Panel on AI, stated that investing in the science and acknowledging what is unknown is more beneficial than instilling fear.

This followed statements from Jacob Coxon, who warned that AI could potentially cause human extinction by the end of the decade. Researcher Miles Brundage, who previously worked at Google DeepMind, echoed similar concerns, suggesting humanity might have already passed the point of avoiding AI-driven eradication. Nobel Peace Prize laureate Maria Ressa emphasized that the discourse must move beyond binary existential threats and "killer robots" narratives to a more nuanced understanding.

The International Scientific Panel on AI, established in 2025, aims to enhance scientific comprehension of AI and guide UN deliberations. Yoshua Bengio, a co-chair of the panel and a key figure in AI research, stressed the need to differentiate between scientists' actual knowledge of AI risks and speculative extrapolations. He highlighted the dangers of inflammatory rhetoric, suggesting that it risks polarizing the conversation.

Bengio called for a global, rational discussion that acknowledges uncertainty and the severity of AI risks.