New Mexico’s AI Teen Decoy Was Evidence Against Snap, Not an Arrest Sting
New Mexico investigators did use AI-generated images to represent a fictional 14-year-old on Snapchat, but the available record places the operation inside a civil investigation of Snap rather than an arrest sting. The Washington Post’s account of the September 2024 complaint identifies the decoy as “Sexy14Heather” and describes recommendations involving accounts seeking sexual exchanges; the materials cited here do not identify anyone arrested or convicted because of contact with the persona.
The lawsuit has advanced since it was filed, although New Mexico has not secured a judgment on its allegations. An April 23, 2025 New Mexico Department of Justice notice states that a court denied Snap’s motion to dismiss and allowed discovery to proceed. Separately, a June 12, 2026 company update says Snap was enforcing against about 220 suspected sextortion instances per day worldwide and submitted approximately 752,000 CyberTips to the National Center for Missing and Exploited Children in 2025.
What the decoy was designed to establish
The fictional profile gave investigators a way to observe what an account presented as a teenage girl might encounter. AI-generated profile pictures made that identity appear credible without putting a real child’s photograph directly into an investigative account.
The profile received recommendations for accounts associated with sexually explicit exchanges, while investigators also examined how Snapchat’s search and recommendation functions responded to the account. That activity became evidence for claims that aspects of the platform’s design could connect young users with adults seeking sexual material.
The documented outcome was evidence against a platform, not a list of captured offenders. Describing the exercise as police using an artificial teenager to “catch sex criminals” collapses two different legal processes: an undercover investigation can document suspicious conduct, but arrest, prosecution and conviction require separate actions and proof against identifiable people.
The distinction also identifies the proper subject of the case. The defendant is Snap Inc., and the dispute concerns product design, consumer representations and child safety. Individual accounts encountered during the operation formed part of the factual background, but the publicly described litigation does not convert every disturbing interaction into an adjudicated crime.
What the synthetic images do—and do not—prove
Using a fictional face can reduce the immediate privacy risk of placing a real minor’s likeness in a hazardous online setting. It may also let investigators maintain a consistent identity without asking an adult officer to pose as a child.
That advantage does not answer every question about the method. The selected public materials do not identify the image generator, the prompts, the model’s training data or the rules governing retention of the resulting pictures. They therefore cannot support claims about whether identifiable children contributed to the model or whether the state used a particular commercial AI system.
The available descriptions also do not establish that investigators generated explicit synthetic child sexual abuse material. They establish a narrower fact: AI-generated profile images represented a purported minor, and the account was then used to observe recommendations and contacts associated with sexual content. Treating those profile pictures as equivalent to generated CSAM would go beyond the disclosed record.
An undercover profile also cannot reproduce every ordinary teenager’s experience. Search terms, accepted contacts and prior interactions can affect what a recommendation system displays. The account can reveal a pathway that deserves examination, but it cannot by itself measure how often that pathway appears across Snapchat or prove what every young user would see.
Why surviving dismissal is not the same as winning
The court’s refusal to dismiss the lawsuit was consequential because it kept New Mexico’s claims alive and opened the way to discovery. That phase can expose internal records, testimony and technical evidence relevant to what the company knew, how its systems operated and whether its public safety representations matched its practices.
A denial of dismissal is nevertheless not a verdict on liability. At that stage, a court determines whether claims may proceed under the applicable legal standards; it does not resolve every disputed fact after a full evidentiary record. The latest official case-specific update cited above leaves the dispute at that intermediate stage.
This procedural status matters when interpreting the decoy operation. The interactions are contested evidence supporting New Mexico’s theory of the case, not a judicial finding that Snapchat’s design caused each alleged harm. A final judgment, settlement or later dispositive ruling would be needed to establish a different outcome.
Snap’s enforcement figures add context, not a resolution
Snap’s 2026 update presents sextortion detection as an ongoing global enforcement task. Its daily figure combines suspected cases found proactively with alleged violations reported by users, and it covers community members of all ages rather than minors alone.
The company’s CyberTip totals are likewise measures of reporting activity, not counts of confirmed offenders, convictions or distinct victims. The 2025 total included about 19,000 tips categorized as sextortion and more than 86,000 involving online enticement or grooming. Those figures may reflect the scale of abusive activity, detection capability, reporting practices or a combination of all three.
They therefore cannot settle New Mexico’s claims about earlier product design and safety representations. They do show why the underlying issue remains current: suspected sextortion continues to generate substantial enforcement and reporting work on Snapchat even after the original lawsuit and the introduction of additional safeguards.
The accurate reading of the case
New Mexico created a synthetic teenage persona to test and document parts of the Snapchat experience, then used the resulting material in civil litigation against Snap. The operation may have exposed accounts and recommendation pathways associated with sexual content, but the cited record does not establish that it culminated in arrests.
The AI element is significant because it changes how investigators can construct online decoys while raising unresolved questions about provenance and oversight. Legally, however, the central conflict remains conventional: New Mexico must prove its allegations about the platform, Snap can contest the evidence and its interpretation, and a case permitted to enter discovery is still an unresolved case.
Also read:
Subscribe to our newsletter
Get the latest Web3, AI, and crypto news delivered straight to your inbox.