Omer Bilgin

We must always endeavour to learn from past mistakes. @caiocvm, @OmerBilginAI and I set out why severe AI incidents warrant policymakers’ attention. The evidence is already in plain sight. Incidents carry information about failure modes, affected populations, and gaps in oversight. But sharing that information is hard, and without it countries have no way to coordinate a response. This is more important now, when AI causes severe harm in many countries at once. We argue that incident sharing is a prerequisite for response to AI harms. Without a way to observe harms systematically and learn across borders, prevention and preparedness have little to work from.