Source: Traverse Legal

Section 1466A of the United States Criminal Code criminalizes knowingly producing, distributing, receiving, or possessing with intent to distribute certain obscene visual depictions of minors engaged in sexually explicit conduct, and the statute expressly covers computer-generated images, drawings, cartoons, paintings, sculptures, and other visual depictions even when no actual minor exists. Congress framed §1466A to address constitutional concerns from Ashcroft v. Free Speech Coalition by limiting the statute to material that also meets the legal obscenity standard, so the absence of a real child does not automatically create a safe harbor for AI-generated images. In United States v. Anderegg (W.D. Wis. 2025), federal prosecutors charged a defendant under §1466A for using Stable Diffusion to produce obscene AI-generated images of fictional minors; the court allowed production and distribution charges to proceed while dismissing only the private-possession charge based on a separate constitutional rule protecting certain private possession inside the home. The piece emphasizes that commercial AI platforms produce and deliver images in response to prompts, conduct that aligns more with production and distribution than with private possession and therefore increases potential criminal exposure.

Operators should begin compliance analysis with an assessment of a model’s capabilities rather than assuming “no real child” ends the inquiry; if a model can produce sexual images of apparent minors, that capability can create federal criminal exposure under §1466A regardless of intent or whether the content is entirely AI-generated. Practical measures recommended include testing prompts, evaluating outputs, identifying prompt-engineering vectors, and building technical controls—prompt filtering, output classifiers, safety tuning, human review, and ongoing testing—into the product so prohibited images are prevented rather than merely prohibited by policy. The text stresses that repeated ability of users to generate prohibited images signals a product-design problem, that contractual restrictions do not eliminate legal risk, and that compliance should be treated as an engineering and product-governance function; the author’s firm offers advisory services to evaluate models, identify compliance gaps, and design safeguards under federal law.