DeepMind safety researcher resigns citing existential AI risk

Leah Libresco has resigned from Google DeepMind’s AGI safety and alignment team, stating she believes artificial intelligence poses an existential threat to humanity. In a statement posted September 15, 2026, Libresco asserted that current development timelines may be insufficient to prevent catastrophic outcomes. She did not allege specific misconduct by DeepMind but framed her departure as a response to broader industry trajectory concerns regarding advanced AI systems. Her resignation adds to a growing list of safety researchers leaving major laboratories amid disagreements over deployment speed and risk mitigation strategies. Google DeepMind has not publicly commented on Libresco’s departure or her stated concerns. The announcement coincides with increased scrutiny of frontier model capabilities and renewed calls for external oversight of alignment research. Industry observers note that such public exits often reflect unresolved tensions between commercial acceleration and precautionary safety mandates within AI organizations.