The UK National Crime Agency and Internet Watch Foundation now want parents to treat a publicly shared photo of a child the same way they would treat leaving a child's address on a public notice board. That sounds extreme. The data says it is not.
Here is what changed. A few years ago, creating a fake explicit image of a specific person required real technical skill. Today, so-called nudification apps, which digitally strip clothing from any uploaded photo, are available in mainstream app stores. A review conducted in January 2026 found 55 such apps on Google Play and 47 in Apple's App Store. These are not fringe tools: they use sign-in systems from Google, Apple, and Discord, and some disguise their charges on credit card statements as innocuous purchases.
For those who want to target a specific child, the tools have grown even more precise. A technique called LoRA fine-tuning lets someone build a custom AI model of a specific person's appearance using as few as 20 photos, in as little as 15 minutes, on consumer-grade hardware. The IWF says criminals are using exactly this approach: scraping photos from public social media and school websites, then generating realistic abuse material without ever contacting the child.
The numbers behind this are hard to absorb. The US National Center for Missing and Exploited Children received 4,700 reports of AI-generated child sexual abuse material in 2023. By the first half of 2025, that figure had reached over 400,000 reports, an average of more than 2,000 per day. The UK's Internet Watch Foundation identified a 260-fold surge in AI-generated abuse videos in 2025 alone. Analysts say it is already impossible to reliably distinguish AI-generated material from real abuse photographs.
Girls are the overwhelming target. Around 94% of identified victims in AI-generated material are female. The problem also shows up inside schools: male students in multiple countries have used these tools against female classmates, feeding publicly available school photos into nudification apps.
The UK government is moving. Legislation going through parliament proposes making it a criminal offence, punishable by up to five years in prison, to possess, create, or distribute AI tools specifically built to generate child abuse imagery. A parallel measure, one of the first of its kind globally, will allow trusted organizations to test AI systems for these vulnerabilities before they are released publicly. Currently, even safety researchers risk criminal liability if they test whether a model can produce such material, which means dangerous outputs often go undetected until offenders find them.
The IWF's own chief technology officer says he is uncomfortable telling parents to limit photo sharing, but sees no other short-term answer. The longer-term answer he wants is AI systems that are built with child safety as a core requirement from the start, not something bolted on after the fact.
For anyone running an organization that posts photos of children, whether a school, a sports club, a nursery, or a company with a public-facing social media account, this guidance has direct operational implications. Consent forms signed even two or three years ago predate the current generation of these tools. They are worth revisiting. The IWF notes that parents have the right to request removal of their child's image from any website or social media account, and that right is worth exercising proactively rather than reactively.