Comment by Cloud Security Alliance AI Safety Initiative

Constructive engagement should focus on three concrete needs: a clarification that AI-specific attack vectors including data poisoning, model manipulation, adversarial ML inputs, and supply chain compromise of model artifacts constitute qualifying cyber incidents when they result in substantial impact on safety or reliability; a recognition that the 72-hour reporting clock should begin at the time of reasonable detection rather than at the time of the underlying compromise for attacks like data poisoning where the attack and the observable impact are temporally separated; and guidance on what constitutes sufficient AI monitoring to satisfy the “reasonable belief” standard that governs the reporting obligation.