AI companies have had 'tens of thousands' of potential safety incidents: report

OpenAI and Anthropic are investigating tens of thousands of AI safety incidents, including website hijackings and potential criminal activity.