Product1 distinct publisher3 min readUpdated
A substitute teacher's sexualized deepfake crossed two schools and three apps. The school could discipline a student; nothing in the escalation path could take the file down.
The Product Desk · Product desk

Compiled by The Product DeskSomething wrong?How this is made
Follow where the image traveled and the reporting route fails differently at every stop. The TikTok post was public and attached to a real name: the account that shared it used the student's own name in its profile, and DeSantiago recognized him from the middle school where he regularly filled in [5]. That is the only version of this a report form is built for, and even then he found it himself, by typing his own name into TikTok search on the evening he reported the incident, which is how he discovered a second student had built an account impersonating him with his name slightly misspelled [7]. Snapchat he heard about secondhand, after the fact [9]. The classroom copy arrived by AirDrop and was revoked before he could do anything with it [11]. Snap's policies do prohibit sharing AI-generated or manipulated content to humiliate or sexually harass someone, and a company spokesperson says there are established processes for reporting it [14], but each of those processes presumes a victim holding a link.
He had assumed at the start that TikTok's guidelines would clear inappropriate photos quickly [3]. What the reporting shows instead is a teacher doing the platform's discovery work and the school's investigative work at the same time: his suspect list came from follower overlap and his own knowledge of which students were friends [6].
The arithmetic is worth stating plainly. Students named or suspected in his account: the one who admitted making the image, the two he inferred from the sharing account's network, the one behind the impersonation profile. Four. Recorded consequences: one in-school suspension [16].
The law is not the missing piece. Fake nude images of nonconsenting adults can amount to cyberstalking and fall under the federal Take It Down Act, which covers computer-generated depictions [13]. But in most of the cases WIRED gathered, the teachers never reported the content to the platforms carrying it [15], and a school disciplinary file is not a deletion request. DeSantiago spent weeks with administrators, the students' parents and the police [19], after an assistant principal told him the school would take care of it [18]. He warned the principal in writing that students unafraid to generate and post images of him were likely circulating material privately in texts and group chats, and that other students could be in it [22].
What the high school offered him was an immediate investigation and the school counselor [20]. He stopped taking middle school assignments until the matter was fully addressed [10], and says he probably will not return to the district this year because he does not feel safe [21]. The district did not respond to WIRED's request for comment [c10b]. By his account, the image kept spreading [9].
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
On March 12, while working his other part-time job at a coffee shop, Luis DeSantiago received texts from a fellow barista saying an AI photo of him was circulating on social media; her cousin wanted to know if it was him.
Luis DeSantiago is a substitute teacher in Los Banos, California.
DeSantiago initially was not worried because he figured inappropriate photos violated TikTok's community guidelines and were usually removed quickly.
Parts of the image resembled a mirror selfie DeSantiago remembered taking at a job in college, but instead of his work uniform he was wearing lingerie and posing suggestively.
One account that shared the image used their real name in their TikTok profile, and it was a student DeSantiago knew from the local middle school where he regularly filled in.
On the evening he reported the incident, DeSantiago searched his name on TikTok and found that a different student had created an account impersonating him, using his name slightly misspelled and the same photo as the profile picture.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Detailed single-publisher account with a named victim, but institutional corroboration is absent
The core narrative is strong for a single source: a named teacher, a dated onset, quoted correspondence with the principal, a specific disciplinary outcome, a platform statement from Snap, and legal framing from a named employment lawyer. It is nonetheless one publisher, one primary witness for the central incident, with no district, school, or police confirmation, no comment from TikTok, anonymous sourcing for corroborating teachers, and no named generation tool. Attribution beyond the student who admitted creating the image rests on the victim's inference.
Multiple concrete incidents across three apps, but no prevalence measurement
Real-world occurrence is demonstrated, not hypothetical: one fully narrated incident spanning TikTok, Snapchat, and AirDrop plus three further teachers describing similar student-made sexualized AI content, and one platform confirming it has policies and reporting processes for exactly this abuse. What is missing is scale — no counts, no district or platform data, no survey — so the pattern is documented as recurring but unquantified, and the remediation side (platform reports, Take It Down Act removals) shows near-zero uptake by the victims themselves.
Slightly overstated: the 'nobody could delete it' framing outruns an untested takedown path
The reported harms and the discipline-versus-removal gap are genuine and well-evidenced. The framing overshoots modestly in two places: the claim that nothing in the escalation chain could remove the file is asserted while the mechanisms designed to do so were largely never invoked — the teachers mostly did not report to platforms, and Snap says established reporting processes exist — and the four-students-versus-one-suspension arithmetic mixes a confirmed admission with unconfirmed suspicion. The direction of the story is supported; its strongest structural claim is under-tested rather than contradicted.
Ordinary reporting and stakeholder incentives, partly visible in the text
Incentives are mild and mostly disclosed. WIRED covers AI-harm narratives that reward vivid single-victim storytelling. The primary witness has an interest in public pressure on a district he says he may not return to and in validating his safety concerns. Snap's on-record statement is reputational positioning toward its own policies. The district's non-response and TikTok's silence mean the parties with the strongest incentive to contest the account are absent, which shapes the balance of the piece. No commercial sponsorship, vendor promotion, or financial stake appears anywhere in the source.
Moderate: incident specifics are solid, systemic and attribution claims are not
Confidence is moderate. The lived facts — the image, the escalation chain, the single suspension, the cross-platform spread, the teacher's withdrawal — are specific, quoted, and internally consistent, and one platform statement is on the record. Confidence drops on everything beyond that one case: single-publisher sourcing, anonymous corroborating teachers, no institutional confirmation, no named tool, and no data establishing how common teacher-targeted deepfakes are.
product
Meta's under-13 data practices go to a jury: 29 AGs, COPPA, and a porous age gate1 distinct publisher
product
A $500 fake cleared Rolex's own London workbench, and staff called it beautiful1 distinct publisher
product
From $50 to $2.25M: Niu Lai prices human authorship, not AI reputational risk2 distinct publishers
product
Cinemas, classrooms and ICE: smart glasses now need a venue-policy contingency1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 24, 2026