A global AI safety assessment noted that traditional evaluation methods struggled to keep pace with rapid advances in general ...