Three Bills Against a Threat the Law Never Named

Texas’s child-protection statutes were written for a slower kind of harm: a camera, a photograph, a manufactured object you could hold in your hand. Generative AI broke that assumption in every direction at once, fabricating images of real children who never posed for anything, manufacturing images of children who don’t exist at all, and turning production into something an app can do in seconds instead of something that requires physical materials. Three bills this session went after three different pieces of that gap.

Texas already criminalized child pornography. It already criminalized obscene devices. What it never had was a specific offense for a manufactured item, a doll, mannequin, or robot, built with a child’s features for sexual use. HB1443 closes that with tiered felonies (state jail up to second-degree) built directly on the Miller v. California obscenity standard, the same constitutional footing Texas’s existing obscenity law already relies on. It passed both chambers without a single no vote.

This is the hardest of the three, because it isn’t one legal question, it’s two. Part of the bill closes an easy gap: using AI to modify or fabricate an image built on a real child’s actual likeness still counts as depicting that real child, squarely inside the settled law from New York v. Ferber (1982). The other part reaches further, criminalizing fully synthetic images of children who don’t exist at all, when the image is realistic enough to be mistaken for a real one. That runs straight into Ashcroft v. Free Speech Coalition (2002), where the Supreme Court struck down a federal ban on virtual child pornography precisely because a fictional depiction isn’t automatically unprotected the way a real child’s image is. SB1621 tries to navigate that by requiring the image be “virtually indistinguishable” from an actual child and requiring the creator to know or believe it depicts a real minor. That’s careful drafting. It’s also genuinely untested law. Alabama’s Child Protection Act, signed in April 2024, uses the same “virtually indistinguishable” standard, and no appellate court has ruled on that standard yet.

Texas already had a narrow deepfake statute covering “videos.” SB441 widens it to any AI-created image, adds a threat offense separate from actual production, and kills the “it’s labeled fake” defense outright. Then it goes somewhere the other two don’t: a civil cause of action that reaches past the person who made the image to the website, app, or “nudification” tool that knowingly profits from facilitating it, or that ignores a 72-hour takedown request. Legitimate AI developers get a real safe harbor if they build in actual technical safeguards, filtering, detection, reporting tools, so the bill targets platforms that profit from the harm, not the technology broadly. Penalties escalate to a third-degree felony when the victim is a minor.

HB1443 named a physical object the law had never covered. SB1621 updated a statute written for photographs to reach images that were never photographs, on legal ground that’s solid for real children and genuinely unsettled for synthetic ones. SB441 reached past the person who made the harm to the platform that profited from spreading it. Three different targets, one shared premise: AI didn’t invent this harm, it just made producing it fast enough that the law had to catch up in a single session instead of a decade.

This is our read of the bills. We encourage you to read them yourself and reach your own conclusions.