The BasePodcastThe Lawfare Podcast48 min

Lawfare Daily: AI, Child Safety, and the Law

In a 48-minute Lawfare episode, Renée DiResta interviews Stanford's Riana Pfefferkorn and Edinburgh's Cyndie Demeocq about research on 71 US federal prosecutions involving AI-generated child sexual abuse material. Pfefferkorn argues existing federal law works; both guests say identifying real victims is the harder problem.

Open the source ↗

Why it matters

The same thing, for different readers.

What it means for you

Making or sharing AI-generated sexual images of children can already bring federal charges in the US, according to the researchers; the idea of a legal loophole is largely mistaken.

For lawmakers

Pfefferkorn argues US federal prosecutors do not need new statutes; the guests point instead to victim identification, provenance, policing of nudify apps and a UK safe harbour for model testing.

For researchers

The study counts only cases that prosecutors chose to bring and that named AI, so it cannot show how common the abuse is or which cases never reached court.

The question it raises

The guests note that image generators are built on real children's photos taken without consent, raising the question of what consent means when a child's picture becomes training data.

In their own words

Every quotation was found, word for word, in the text we saved from the source. That shows the words are there; it does not make them true, or prove the speaker meant them as quoted.

From our own transcript of the episode’s audio, made by machine. Speaker names come from what is said, not from a voice match.

“We're seeing prosecutions work just fine. What I do think is worthy of closer scrutiny is the impact of AI on victim identification”
Riana Pfefferkorn, Policy fellow, Stanford Institute for Human-Centered Artificial IntelligenceAnswering whether AI has made prosecutions impossible, as Justice Thomas once predicted; the speaker discusses the paper and the Anderegg case.
Found word for word in the sourceAlso in a second, independent transcriptSpeaker known from context, not confirmed by listening
“Everybody else who even tried to bring a First Amendment challenge did not succeed, and that's a total of eight cases out of more than seventy”
Riana Pfefferkorn, Policy fellow, Stanford Institute for Human-Centered Artificial IntelligenceDescribing the study's sample of federal prosecutions and how rarely First Amendment challenges were raised or succeeded.
Found word for word in the sourceAlso in a second, independent transcriptSpeaker known from context, not confirmed by listening
“that is the biggest challenge potentially for law enforcement is not prosecuting somebody because at least in the US my paper shows that our laws are adequate, at least at the federal level, it's the victim identification component”
Riana Pfefferkorn, Policy fellow, Stanford Institute for Human-Centered Artificial IntelligenceSaying the biggest challenge for law enforcement is not prosecuting offenders but telling whether an image shows a real child in need of rescue.
Found word for word in the sourceAlso in a second, independent transcriptSpeaker known from context, not confirmed by listeningspoken clean-up: that (said twice)

How the finding was reached

In our words, from the source. Weigh the finding by how it was made.

What was done
Review of 71 US federal criminal prosecutions, located through federal court records and Department of Justice press releases, whose charging documents expressly mentioned AI; cases from 2026 were excluded as too early.
What it found
The guests say 59 of the 71 cases ended in guilty pleas and 6 in convictions at trial, so 65 convictions; only 8 cases raised First Amendment challenges, and only part of one charge in United States v. Anderegg succeeded. Actual abuse material was alleged in 55 cases, nudify apps in 32 and fully synthetic imagery alone in 9.
What it does not show
Pfefferkorn says the sample covers only cases prosecutors chose to bring and that named AI, so it cannot show what was not prosecuted or how widespread the abuse is.
Status
A working paper: not yet peer reviewed

The record

Questions
How is AI changing childhood, health and daily life? · Who can demand evidence, delay a release or stop an unsafe deployment?
People
Renée DiResta (Speaks, Lawfare contributing editor) · Riana Pfefferkorn (Speaks, Policy fellow, Stanford Institute for Human-Centered Artificial Intelligence) · Cyndie Demeocq (Speaks, Doctoral researcher, University of Edinburgh School of Informatics)
What we read
We read the whole source.
Found by
The Radar, run of October 23, 2026, via pod-the-lawfare-podcast (tier 1: the source’s own channel or an official record)
Cite this
The Lawfare Podcast, “Lawfare Daily: AI, Child Safety, and the Law”, October 22, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/lawfare-2026-10-22-ai-child-safety-law
Our notesSpeech-to-text with no speaker labels and some garbled words (names, 'Anderegg'); who is speaking is inferred from introductions and content, hence speakerCertain false. The paper is described as coming out next year; its publication status was not checked. The episode discusses child sexual abuse material.