Parents in the UK are being told to stop posting their children’s photos online. Schools are removing pupils’ faces from their websites. The National Crime Agency (NCA) has warned that criminals are using publicly available images of children to create AI-generated child sexual abuse material, and that most parents have no idea this is happening.
If the NCA’s advice applies to a family snapshot on Instagram, it applies to a campaign image of a girl in Kigali or Accra on a charity website. The AI tools that turned a school photograph in Loughborough, UK into abuse material do not distinguish between a pupil in the UK and a child in a low-income country. This conversation now extends beyond ethics, representation, or even consent. It is about safeguarding.
What the evidence now shows
The scale of the threat has moved beyond what most organisations have grasped.
The Internet Watch Foundation (IWF) identified more than 8,000 AI-generated images and videos depicting realistic child sexual abuse in 2025. Girls accounted for 97% of the victims. AI-generated child sexual abuse videos saw a 26,385% increase in a single year, and 65% of those videos were classified as Category A, the most severe legal classification under UK law.
Earlier this year, reporting revealed that criminal gangs had targeted a UK secondary school by scraping photographs of pupils from the school’s website, using AI to manipulate those images into child sexual abuse material, and then attempting to extort the school. The IWF confirmed that approximately 150 of the manipulated images could be classified as criminal content under UK law. The Loughborough Schools Foundation has since redesigned its website to remove recognisable images of pupils.
The risk also operates at a systemic level. Human Rights Watch found that the LAION dataset, one of the large-scale datasets used to train popular AI image generators, contained personal photographs of more than 700 Australian and Brazilian children scraped without their consent. Those images were used to train AI models that were then used to create sexually explicit deepfakes of other children. Human Rights Watch reviewed less than 0.0001 percent of the dataset, meaning the real number of children affected is almost certainly far higher. Researchers have confirmed that identifiable details such as names, locations, and timestamps heighten the risk of misuse, and that children face particular vulnerability to lasting harm from privacy breaches of this kind.
This is significant harm to a child and it begins with a photograph being published online.
We cannot look away
Schools across the UK are removing names, reducing identifying information, and reconsidering what remains accessible on their websites.
Yet many development organisations continue to publish exactly the kind of material that UK authorities are urging families to reconsider. The images and information we publish sit on public-facing websites indexed by search engines, in downloadable annual reports, in fundraising emails sent to tens of thousands of recipients, and across social media platforms where they are designed to be shared widely. Many remain accessible for years.
The children in these images are overwhelmingly in low-income countries and in contexts where there is no equivalent of the NCA, no GDPR framework, and no mechanism for redress if their images are misused. The question of consent runs deeper than many organisations have been willing to acknowledge. It is not only that power dynamics between an international NGO and a child living in poverty or displacement make free and equal consent difficult. It is also that consent processes themselves are often inadequate, failing to give children or families meaningful information about how images will be used, for how long, or across which platforms. And at a more fundamental level, children cannot meaningfully consent to being used in fundraising because they do not understand what fundraising is, how their image will function within it, or what the long-term consequences of that exposure might be.
The ethical case for change has been made repeatedly. Grace Gatera, a lived experience mental health advocate in Kigali, told us: “They take pictures of our pain.” What has changed is that it is now also a child protection case, because the evidence from the NCA, the IWF, and Human Rights Watch tells us that publishing identifiable images of children online carries foreseeable risks of serious harm that did not exist when many organisational communications policies were written.
This does not mean every historic decision was wrong but it does mean every organisation should ask whether its current practice reflects today’s reality.
Duty of care is about anticipating foreseeable risks and taking reasonable steps to reduce them, and in an AI-enabled world, that responsibility now extends to every image an organisation places in the public domain.
A different approach is possible
In March this year, Chance for Childhood hosted a webinar with academic researcher Arsenii Alenichev, whose work has revealed how AI image generators default to stereotypical and harmful depictions of people in African and Asian countries, from white-saviour imagery to the erasure of Black professionals. The resulting blog explored these risks alongside the finding that generative AI models are trained on faces scraped without consent. That piece focused on what happens when AI generates images of the communities charities serve. The question now is what happens when AI takes the images charities have already published.
At Chance for Childhood, we chose to stop showing identifiable images of children when we launched the #OverExposed campaign in 2022. That decision was grounded in rights, dignity, and the recognition that the children we serve deserve the same protections as children anywhere. The NCA’s warnings underscore the urgency of that decision, and we are continuing to learn and develop our practice as the landscape changes.
Changing course does not mean organisations stop telling stories, and it does not mean fundraising becomes impossible. The alternative could be photographs taken from a distance, images that do not show faces, illustrations, and stories told in partnership with communities rather than extracted from them. These approaches are available to any organisation willing to invest in doing this differently.
The sector has now been warned by these findings of the UK’s own law enforcement agencies, who have told us that the images we publish carry risks we can no longer ignore. The question is how we respond to this warning. At Chance for Childhood, we are inviting NGOs, communicators, fundraisers, filmmakers, photographers, and anyone who tells stories about children in the development sector to join us on that journey through the #OverExposed campaign.
Visit www.chanceforchildhood.org/overexposed to learn more, explore our resources, and take the pledge.
Read our blog on AI and charity imagery https://www.chanceforchildhood.org/ai-and-charity-imagery/
Sources
IWF, “Harm without limits: AI child sexual abuse material through the eyes of our Analysts”: https://www.iwf.org.uk/about-us/why-we-exist/our-research/how-ai-is-being-abused-to-create-child-sexual-abuse-imagery/
IWF, 2025 Annual Data and Insights Report: https://www.iwf.org.uk/annual-data-insights-report-2025/
IWF, AI-generated images data page: https://www.iwf.org.uk/annual-data-insights-report-2025/image-insights/ai-generated-images/
Human Rights Watch, “720 Australian and Brazilian Children Better Protected from AI Misuse”: https://www.hrw.org/news/2024/09/03/720-australian-and-brazilian-children-better-protected-ai-misuse/
Human Rights Watch, Australia report on children’s photos in AI training data: https://www.hrw.org/news/2024/07/03/australia-childrens-personal-photos-misused-power-ai-tools
The Guardian, “AI sexual abuse fears: UK parents warned over posting images of children,” 3 July 2026: https://www.theguardian.com/society/2026/jul/03/ai-sexual-abuse-fears-uk-parents-warned-posting-images-children-national-crime-agency
Malwarebytes/Security Boulevard on schools removing student photos: https://securityboulevard.com/2026/05/deepfake-sextortion-forces-schools-to-remove-student-photos-from-websites
UK Safer Internet Centre Schools given advice on image safety to keep ahead of threat from AI blackmailers – UK Safer Internet Centre
Chance for Childhood https://www.chanceforchildhood.org/ai-and-charity-imagery/