We don’t imprison humans preemptively based on the capability to commit crime. Why regulate AI that way?
The Trump administration is reportedly considering a regulatory approach for AI that focuses on pre-deployment testing and capability assessment, a move criticized for mirroring outdated consumer product regulations. Experts argue this ex ante approach is flawed because AI systems' behavior is unpredictable and evolves, similar to human actions. Instead, they advocate for an AI Safety Management System centered on continuous, real-world evaluation and accountability for actual harm, rather than hypothetical capabilities. AI
IMPACT Focus on post-deployment accountability could shift AI development priorities towards robust real-world safety and away from purely capability-based benchmarks.