The National Crime Agency and the Internet Watch Foundation delivered a stark message this week. Parents should keep photos of their children off public view on social media. The reason sits in the numbers. And those numbers have become terrifying.
In 2025 the IWF assessed 8,029 AI-generated images and videos of realistic child sexual abuse material. That figure marked a 14 percent increase from the year before. The real shock came in the videos. The organization found 3,443 of them. Compare that to 13 in 2024. The jump reached 26,385 percent. Sixty-five percent of those videos fell into Category A. They showed the most extreme acts.
But. These aren’t abstract threats. Criminals scrape ordinary family snapshots from Instagram, Facebook, school websites. They feed them into generative tools. The output looks photographic. It moves like video. No grooming required. No physical access. Just public pixels and a few prompts.
The BBC reported the joint warning on July 4, 2026. Tim Wright, senior manager at the NCA, said parents and carers should take a few simple steps today. Kerry Smith at the IWF put it bluntly. These are not hypothetical threats. They are real.
The new parent guide from the IWF makes three recommendations. Check privacy settings on every account. Review who can see images of your children. Discuss consent openly when schools or clubs ask for photo permission. Audit old posts too. That school-uniform shot from three years ago? It can still be scraped. Make accounts private. Use close-friends lists. Withdraw consent where possible.
Lorna Sinclair, child sexual abuse education manager at the NCA, captured the gap in awareness. The average parent or carer does not post a picture of a child online thinking that it might be scraped to be turned into CSAM. There are lots of parents and carers who do not know that this problem exists.
Dan Sexton, chief technology officer at the IWF, sounded exhausted in The Guardian’s coverage. I don’t know what else to say to parents. I would be very cautious about putting pictures of children online because there is no protection. He admitted discomfort at issuing the advice. Yet he saw no alternative.
The scale extends far beyond Britain. NCMEC in the United States tracked the same surge. Reports of AI-linked child sexual exploitation hit roughly 4,700 in 2023. They climbed to 67,000 in 2024. By 2025 the total surpassed 1.5 million. More than 275 direct victims were identified across 2024 and 2025. Many offenders already knew the children. A scraped innocent photo from a parent’s feed became the seed.
The IWF’s own research page details the technological advance. Tools have converged. One application now generates imagery with minimal effort. LoRAs let users fine-tune models with as few as 20 images in 15 minutes. Girls make up 97 percent of the victims depicted. The material grows more extreme. Full-motion video has moved from dark-web rarity to commonplace. Chatbots simulate abuse scenarios on the open web. Audio deepfakes clone voices. The barrier to entry has collapsed.
Earlier IWF data showed a 380 percent rise in actionable AI reports in 2024. That year marked the shift from static deepfakes to the first realistic videos. The 2025 explosion confirmed the trend. Harms spread to both dark web forums and mainstream commercial platforms. Some images mimic amateur photography to dodge detection. Others train on real abuse material. They re-victimize survivors while normalizing violence against children.
Real cases illustrate the danger. UK schools faced blackmail after criminals scraped pupil photos, generated abuse imagery, and threatened to release it. The IWF heard from under-18s extorted after AI nudified their clothed selfies. One 15-year-old girl described a convincing fake nude that used her face and her own bedroom background lifted from Instagram.
UNICEF warned in a February 2026 brief that generative tools mark a significant escalation. Across 11 countries in one project, at least 1.2 million children reported images manipulated into explicit deepfakes. In some places that equals one child per classroom. The agency called for training among parents, educators, and law enforcement.
Stanford HAI examined school responses to student misuse of nudify apps. West Virginia passed legislation after experts highlighted family risks. The FBI reiterated years ago that even fully artificial CSAM remains illegal under federal law. Production, distribution, possession. All prohibited.
Yet enforcement lags. Platforms face pressure but detection tools struggle with volume and sophistication. The IWF noted that without direct confrontation the problem will only continue to grow. Infinite violations become possible with unprecedented ease.
So what now? The NCA and IWF avoid telling parents never to photograph their children. They stress awareness. Delete what you can. Limit visibility. Talk to family members who share your kids’ images. Revisit every consent form signed before AI changed the equation.
Tom Dyson, IWF head of marketing, reminded parents they hold power. If you want a photograph of your children to be taken off a website or social media, you are perfectly able to do that.
The guidance arrives at a moment when many families still treat social media as a casual photo album. That habit now carries measurable risk. One innocent ballet recital post can fuel a criminal’s catalog. One birthday party video can train a model. The technology no longer needs a dark room or specialized skills. A smartphone and free tools suffice.
Industry insiders tracking online safety have watched this curve steepen for three years. What began as experimental deepfakes has matured into production-grade abuse engines. The latest reports show not just growth but acceleration. Videos multiplied by orders of magnitude. Quality improved enough to fool the untrained eye. And the victims remain overwhelmingly real children whose faces were never meant for this.
Parents cannot outrun every scraper. But they can reduce the supply of fresh, high-quality source material. Private accounts. Restricted sharing. Conscious consent. These steps won’t solve the broader problem. They do give families a defensive layer while regulators, platforms, and technologists race to catch up.
The message from UK authorities carries weight because it comes from the front line of removal and prosecution. They see the reports daily. They analyze the imagery. They talk to the victims. When they say the threat is real and growing, the data backs them. The question is whether enough parents will hear the warning before their own child’s image joins the torrent.
WebProNews is an iEntry Publication