On February 4, 2026, UNICEF issued a statement from New York under a blunt headline: “Deepfake abuse is abuse.” The agency called on governments to expand the legal definition of child sexual abuse material (CSAM) to cover AI-generated imagery, and to criminalize its creation, procurement, possession, and distribution. Reuters covered the call that week, and UNICEF USA repeated the message in a Forbes essay timed to Safer Internet Day on February 10.
This is not routine advocacy. It arrives with the largest child-facing survey evidence to date: at least 1.2 million children reported in the past year that images of them had been manipulated into sexualized deepfakes. For the AI industry, the statement reframes platform responsibility from after-the-fact removal to pre-publication blocking, and it points directly at the guardrails of image-generation models.
Where the 1.2 Million Figure Comes From
The number comes from Disrupting Harm Phase 2, a research program led by UNICEF’s Office of Strategy and Evidence – Innocenti with ECPAT International and INTERPOL, funded by Safe Online. Surveys ran in 11 countries, each polling roughly 1,000 internet-using children aged 12 to 17 plus about 1,000 parents or caregivers, with national coverage rates between 91 and 100 percent. Estimates were weighted using UN 2024 population data and estimated child internet-use rates.
Three numbers stand out. First, at least 1.2 million children disclosed that images of them had been turned into sexualized deepfakes within the past year. Second, in some countries that works out to one child in 25 — UNICEF’s shorthand is one child in a typical classroom. Third, in some countries up to two-thirds of children said they worry AI could be used to create fake sexual images or videos of them. National reports with country-level findings will be released through 2026.
Three Asks: Governments, Developers, Platforms
The statement assigns homework to three audiences. Governments: expand CSAM definitions to cover AI-generated content, and criminalize its creation, procurement, possession, and distribution — including possession is a stricter standard than many current laws set. AI developers: adopt safety-by-design approaches and robust guardrails so models cannot be misused to generate abusive material. Digital companies: stop circulation proactively instead of removing material only after abuse occurs, and invest in detection so content comes down immediately, not days after a victim’s report.
The package also references UNICEF’s Guidance on AI and Children 3.0, updated in December 2025, and an issue brief on AI and child sexual abuse and exploitation, which frames generative tools as a significant escalation of risk to children.
The Definition Fight: Closing the “It’s Fake” Defense
The sharpest line in the statement does definitional work: “Sexualised images of children generated or manipulated using AI tools are child sexual abuse material.” That sentence closes a legal gray zone. As long as an image is classified as synthetic, offenders and platforms can argue there is no real victim. UNICEF’s position is that the harm does not depend on where the pixels came from — “Deepfake abuse is abuse, and there is nothing fake about the harm it causes.”
The statement also names “nudification” tools: applications that use AI to strip or alter clothing in photos and fabricate nude images. These tools are cheap to build and spread fast, and they are the main technical vehicle behind the 1.2 million figure. The closing warning left for legislators: “Children cannot wait for the law to catch up.”
What It Means for Builders and Product Teams
For image-generation and editing products, the statement previews where regulation is heading. Safety-by-design stops being a model-card pledge and becomes a statutory expectation: on the input side, requests to modify photos of real children must be blocked; on the output side, generated media must be detectable and traceable. The proactive-detection demand pushes classifiers and watermarking from nice-to-have into infrastructure — part of the safety and regulation theme we sketched in our opening outlook for 2026.
The more immediate effect is a shift in liability. Once “removed days later” is publicly named as unacceptable, review SLAs, moderation pipelines, and model guardrails all face sharper scrutiny. Teams building content-safety pipelines should fold child-protection scenarios into red-teaming and evaluation metrics now; retrofitting after legislation lands will cost far more than building it in.
Sources
- Deepfake abuse is abuse — UNICEF
- UNICEF calls for criminalization of AI content depicting child sex abuse — Reuters
- UNICEF Urges Action to Protect Children from AI-Generated Sexual Content — Forbes
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.