A team of MIT scientists, led by graduate student Vinith Suriyakumar and associate professors Ashia Wilson and Marzyeh Ghassemi, joined forces with researchers from Thorn to develop a new auditing approach that determines whether a model can produce child sexual abuse material (CSAM) without prompting it. The National Center for Missing and Exploited Children received more than 1.5 million reports of AI-generated CSAM in 2025, an increase from 67,000 in 2024. The technique examines hidden representations and never generates an output. When tested, the auditing procedure identified model variations that had been specialized to generate CSAM with 100 percent accuracy. Suriyakumar stated the method unlocks a new avenue for platforms and law enforcement to test whether a model is capable of generating CSAM, addressing an AI safety problem with severe negative impacts. The paper was presented as a spotlight at the "Trustworthy AI for Good" workshop at the International Conference on Machine Learning.