Risks likely increase if AI development happens quickly and in an uncoordinated way across many groups. International cooperation on safety standards could help, but may be difficult to achieve.