Elon Musk’s AI chatbot has come under fire after reports that it can be manipulated to generate sexually explicit deepfake images, including images involving minors. The controversy has renewed concerns about how artificial intelligence systems are tested and regulated before they are released to the public.
Riana Pfefferkorn, a former technology policy lawyer, argues that the most effective way to combat AI-generated sexual abuse material is to prevent it at the source. In her view, that means allowing developers and independent researchers to rigorously test AI models to identify and close off ways they can be misused—before harmful content appears online.
However, Pfefferkorn warns that current U.S. laws do not clearly protect researchers who conduct this kind of good-faith testing. As a result, companies may avoid probing their own systems too deeply out of fear that testing could expose them to criminal liability. The law, she argues, fails to clearly distinguish between responsible testers and bad actors.
Pfefferkorn calls on Congress to pass a national law creating a legal safe harbor for AI developers and researchers who responsibly test models for vulnerabilities related to child sexual abuse material. She notes that a similar delay occurred in cybersecurity, where legal protections for ethical hackers were not established until after major foreign cyberattacks prompted action in 2022.
“We cannot afford years of government inaction again,” Pfefferkorn writes. She urges lawmakers to hold immediate hearings on the issues raised by Musk’s Grok chatbot and to move quickly on legislation. While companies such as Musk’s xAI could take stronger steps to improve safety, she argues that clear legal protections are essential—and time is running out.