MIT's New Method Flags AI Models Trained on CASM Without Generating It

MIT researchers have developed "Gaussian probing," a technique that can identify AI models fine-tuned to generate child sexual abuse material with 100% accuracy without ever producing an illegal image. The method analyzes internal model adaptations rather than outputs, bypassing the legal and ethical barriers that previously made safety audits impossible. This breakthrough addresses a critical security gap as reports of AI-generated CSAM have surged from 67,000 in 2024 to over 1.5 million in 2025.