Blog Post | 115 KY. L. J. ONLINE | November 8, 2026
United States v. Anderegg: Balancing the First Amendment with Child Safety in the Age of AI
By: Jacob Levy, Staff Editor, Vol. 115
On August 25, 2026, the U.S. Seventh Circuit Court of Appeals, in United States v. Anderegg, issued an opinion regarding the use of generative artificial intelligence (AI).[1] Specifically, the court held that possession of AI generated images containing child sexual abuse material (CSAM) is protected under the First Amendment as applied, and therefore a defendant cannot be convicted for production, distribution, or possession of CSAM that is AI generated under existing statutes.[2] Reactions from legal experts and members of the media alike have been highly critical of this holding.[3] The Seventh Circuit’s holding in Anderegg fails to conform with existing precedent[4] and creates significant public policy concerns for child safety by enabling internet predators.
I. How the Seventh Circuit Failed to Conform with Existing Precedent
In Anderegg, the defendant was charged in the Western District of Wisconsin with four counts of producing, distributing, transferring to minors, and possessing CSAM under 18 U.S.C § 1446A(b)(1) and (b)(4).[5] Anderegg was alleged to have sent direct messages containing AI generated CSAM to an account belonging to a minor on Instagram.[6] In his messages, Anderegg explained how he used the generative AI program, Stable Diffusion, to create the explicit images.[7] The Western District of Wisconsin granted Anderegg’s motion to dismiss the possession charge on the grounds that the generated images depict fictional children and are therefore protected under the First Amendment, and the Seventh Circuit upheld.[8] In other words: The Seventh Circuit’s holding protects AI generated CSAM under First Amendment if the child is fictional.[9]
The Seventh Circuit misapplied multiple relevant precedents. First the majority relied upon Ashcroft v. Free Speech Coalition, and Judge Lee’s concurring opinion acknowledges the public policy drawbacks of relying on such an outdated holding: “‘in 2002, the Court's reasoning [in Free Speech Coalition] made sense: Technology was far less developed than it is today.’”[10] The defense relied upon Stanley v. Georgia, arguing that the First Amendment’s protection of possessing obscene materials in one’s own home should apply to all obscene materials.[11] But, in United States v. Andersson the Seventh Circuit refused to extend the Stanley protection to CSAM because Stanley specifically protected obscene images of consenting adults.[12] In Osborne v. Ohio, the Supreme Court held that a statute prohibiting the possession and consumption of CSAM was valid under the First Amendment because such a prohibition serves a legitimate governmental interest in protecting the physical and mental wellbeing of children.[13] The Seventh Circuit considered Osborne in its analysis, however it did not find it controlling because in Ashcroft, the Supreme Court said that there is no compelling government interest in protecting the wellbeing of children when those children are fictional.[14] This fails to consider Free Speech Coalition v. Paxton, where the Supreme Court upheld a statute requiring age verification for all pornographic websites, because the government’s compelling interest in preventing children from accessing pornography outweighs any First Amendment concerns.[15] Finally, the Seventh Circuit failed to correctly apply the third prong of the obscenity test from Miller v. California; Anderegg’s AI generated fictional CSAM “taken as a whole, lacks serious literary, artistic, political, or scientific value.”[16]
II. How this Opinion Endangers Children & Enables Predators
The defense’s reliance on Stanley hinges on the idea that Anderegg’s use of AI to generate CSAM occurred in the privacy of his home.[17] But, this argument fundamentally misunderstands how generative AI works; while the program that the defendant used in this case, Stable Diffusion, can generate images without using the internet, all AI models require training from existing images and knowledge to generate content.[18] This means, regardless of whether the defendant trained the AI model himself entirely within his own home, the AI model was still required to be trained with real images to generate CSAM.[19] This leads into the first public policy concern from this holding: Generative AI requires training on real images of children.[20] Even AI generated images of fictional characters must take inspiration from real images.[21] The defense argued that because the images generated were not images of a real child and occurred entirely within the defendant’s own home, the government’s compelling interest in protecting the victims of CSAM did not exist.[22] This assertion disregards the fact that, at minimum, the defendant would be required to train the AI model on an existing image of a child, and more than likely used real CSAM to train his AI model.[23]
AI generated CSAM also poses significant danger for the victims and potential victims of child sexual abuse. AI is getting better at generating hyper-realistic images every day, which allows child predators to create CSAM without ever accessing a child.[24] AI platforms allow any child predator to take a perfectly innocent photo of a child, turn it into a hyper-realistic sexually explicit photo of a ‘fictional’ character, and, under the Seventh Circuit’s holding in Anderegg, face no legal consequences.[25] These images are used by child predators to harm their victims in one of two ways: (1) to groom victims by normalizing sexual behavior to children who cannot consent to it, as was the case in Anderegg;[26] or (2) spreading the images around the internet, utilizing them as blackmail against children and their families.[27] Furthermore, Anderegg has placed a significant hurdle for law enforcement in preventing child sexual abuse through AI generated CSAM because intervention can only occur after an actual, not virtual, child has already been harmed.[28]
Finally, it should be noted that the concurring opinion in Anderegg discusses many of these public policy concerns.[29] The concurrence acknowledged that due to the significant advances in generative AI and the potential for harm to children, that this case was difficult to decide.[30] As a result, they requested guidance from the Supreme Court, who should consider the precedent and public policy concerns discussed above, and reverse this decision.[31] Judge Lee ended his concurrence by saying: “‘[t]he prevention of sexual exploitation and abuse of children constitutes a government objective of surpassing importance.’”[32] That wisdom continues to ring true today, and it is why the Supreme Court must overturn this decision.
[1] United States v. Anderegg, No. 25-1354, slip op. at 3 (7th Cir. August 25, 2026) (majority opinion), https://media.ca7.uscourts.gov/cgi-bin/OpinionsWeb/processWebInputExternal.pl?Submit=Display&Path=Y2026/D08-25/C:25-1354:J:Lee:autconi:T:fnOp:N:3597567:S:0.
[2] See id. at 2–3; 18 U.S.C § 1466A(b)(1) (2003) (“Any person who… knowingly possesses a visual depiction of any kind, including a drawing, cartoon, sculpture, or painting, that… depicts a minor engaging in sexually explicit conduct; and is obscene… shall be subject to the penalties provided in section 2252A(b)(2).”).
[3] See, e.g., Nitasha Tiku, Federal Judge Warns Law is Being Left Behind by AI Child Sex Abuse Images, Wash. Post (Aug. 26, 2026), https://www.washingtonpost.com/technology/2026/08/26/federal-judge-warns-law-is-being-left-behind-by-ai-sex-abuse-images.
[4] See generally Stanley v. Georgia, 394 U.S. 557, 568 (1969) (providing that the right to possess obscene material within one’s home is protected under the First Amendment); United States v. Andersson, 803 F.2d 903, 906–07 n.3 (7th Cir. 1986) (holding that the protection in Stanley does not extend to obscene material that depicts minors); Osborne v. Ohio, 495 U.S. 103, 109, 111 (1990) (holding that preventing the consumption and production of CSAM serves a legitimate governmental interest); Free Speech Coal. v. Paxton, 606 U.S. 461, 466–68, 472 (2025) (“History, tradition, and precedent recognize that States have two distinct powers to address obscenity: They may proscribe outright speech that is obscene to the public at large, and they may prevent children from accessing speech that is obscene to children.”); Miller v. California, 413 U.S. 15, 24 (1973) (holding that obscene material may be censored by the government if “taken as a whole, does not have serious literary, artistic, political, or scientific value.”).
[5] See Anderegg, slip op at 4–5 (alteration in original) (quoting 18 U.S.C § 1466A(b)(1) (2003)) (explaining that “Anderegg was charged with (1) ‘knowingly produc[ing]’ and (2) ‘knowingly distribut[ing]’ visual depictions of minors engaged in sexually explicit conduct as well as (3) ‘knowingly transferr[ing]’ such material to a minor under the age of sixteen” and “‘knowingly possess[ing] at least one visual depiction that depicted a minor engaging in sexually explicit conduct and was obscene, and any depiction involved in the offense had been shipped and transported in interstate and foreign commerce’”)).
[6] Id. at 3.
[7] See id. at 4.
[8] Id. at 2–3, 5.
[9] See id. at 3.
[10] Id. at 25 (Lee, J., concurring) (quoting Jessica L. Terkovich, The Emerging Danger of AI-Generated Child Sexual Abuse Material and an Unprepared Crim. Code, 45 N. Ill. U. L. Rev. 398, 404 (2025)).
[11] See id. at 18 (majority opinion) (citing Stanley v. Georgia, 394 U.S. 557, 568 (1969)).
[12] United States v. Andersson, 803 F.2d 903, 906–07 n.3 (7th Cir. 1986) (citing Stanley v. Georgia, 394 U.S. 557 (1969) (explaining that the defendant did not have the right to possess obscene material of children within his own home because “the State's interests in regulating child pornography are radically different than those in regulating obscenity” and even if he did, “that right in no way extends to acquiring or providing such material for private use.”).
[13] Osborne v. Ohio, 495 U.S. 103, 109, 111 (1990) (“The State does not rely on a paternalistic interest in regulating Osborne's mind. Rather, Ohio has enacted [the statute] in order to protect the victims of child pornography; it hopes to destroy a market for the exploitative use of children.”).
[14] See Anderegg, slip op. at 12 (citing Ashcroft v. Free Speech Coal., 535 U.S. 234, 250 (2002)); Osborne v. Ohio, 495 U.S. 103, 109 (1990)).
[15] Free Speech Coal. v. Paxton, 606 U.S. 461, 466–68, 472 (2025) (“History, tradition, and precedent recognize that States have two distinct powers to address obscenity: They may proscribe outright speech that is obscene to the public at large, and they may prevent children from accessing speech that is obscene to children.”).
[16] Miller v. California, 413 U.S. 15, 24 (1973).
[17] Anderegg, slip op. at 18 (citing Stanley v. Georgia, 394 U.S. 557, 568 (1969)).
[18] Andrew Zhu, How does Stable Diffusion work?, Stable Diffusion Art, https://stable-diffusion-art.com/how-stable-diffusion-work/#Ethical_moral_and_legal_issues (last visited Sep 13, 2026).
[19] Id.
[20] Id.
[21] Olivia D’Andrea, Synthetic Abuse, Real Consequences: Confronting the Challenge of AI-Generated CSAM, Nat’l Ass’n of Att’ys Gen., https://www.naag.org/attorney-general-journal/synthetic-abuse-real-consequences-confronting-the-challenge-of-ai-generated-csam (last visited Sep. 13, 2026).
[22] United States v. Anderegg, No. 25-1354, slip op. at 19 (7th Cir. August 25, 2026) (majority opinion), https://media.ca7.uscourts.gov/cgi-bin/OpinionsWeb/processWebInputExternal.pl?Submit=Display&Path=Y2026/D08-25/C:25-1354:J:Lee:autconi:T:fnOp:N:3597567:S:0.
[23] See Benjamin L.W. Sobel, A Real Account of Deep Fakes, 124 Mich. L. Rev. 871, 930 (2026).
[24] See D’Andrea, supra, note 20.
[25] Anderegg, slip op. at 16.
[26] Id. at 3–4.
[27] Press Release, U.S. Secret Serv., Sextortion in the Age of AI, (Aug. 12, 2026) (available at https://www.secretservice.gov/newsroom/behind-the-shades/2026/08/sextortion-age-ai).
[28] Anderegg, slip op. at 26 (Lee, J., concurring) (citing Addressing Real Harm Done by Deepfakes: Hearing Before the Subcomm. on Cybersecurity, Info. Tech., and Gov’t Innovation of the H. Comm. on Oversight and Accountability, 118th Cong. 6 (2024) (testimony of John Shehan, Vice President, Exploited Children Division, Nat’l Ctr. For Missing & Exploited Child.)).
[29] Id. at 25.
[30] Id. at 27.
[31] Id. at 25.
[32] Id. at 27 (alteration in original) (quoting: New York v. Ferber, 458 U.S. 747, 757 (1982)).