The Internet Watch Foundation (IWF) publishes a report on the scale of AI child sexual abuse material (CSAM) in 2025.
The key AI models that are of concern to the IWF are large language models (LLMs), text-to-image models and ‘nudify’ (using AI to generate nude images from clothed photos) apps. Not only are these concerning because of the ability to generate artificial CSAM, but the IWF raises concerns about extended harms. For example, the risk of AI companion chatbots grooming children:
“When children interact with companion chatbots that steer them towards sexual conversations, the relational dynamics can become reminiscent of established grooming patterns. This risk is heightened by the “sycophantic” communication style of many chatbots, which validate and flatter the user. Children risk being exposed to sexual topics, and encouraged to prolong engagement.”
Key findings
Other key findings include:
- In 2025, the IWF identified 3,443 AI-generated child sexual abuse videos – a 26,385% increase from the previous year. And 65% of these videos are of the most extreme classification – category A.
- Girls are overwhelming targeted, with 97% of illegal AI generated imagery assessed in 2025 depicting girls, reflecting on the issue of persistent, gendered sexual violence online and cultures that normalise misogyny and the humiliation of girls.
- Chatbots are being used to share AI-generated sexual abuse images and ‘encouraging users to act out simulated child sexual abuse scenarios’.
- With advances in AI comes easier generation of these materials – tools have more capabilities in a single application.
- This advancement also amplifies the realism of content and therefore its potential harm.
- Open source models (code publicly available to download, customise and fine tune) make it easier to create models that are trained on CSAM and the gap in sophistication between open and closed source models is narrowing.
- A substantial portion of AI CSAM isn’t visible to the IWF – this is because some of these models and datasets are being hosted offline and shared via end-to-end encrypted peer networks. So it’s hard to know the full scope of just how widespread this material really is.
Artificial material, real harms
The amount of AI-generated material is small when compared with the amount of material the IWF acts on each year but that does not minimise its harm. As the IWF explains:
“AI-generated sexual imagery of children causes real and lasting harm to victims – even if the physical abuse depicted did not take place. From a victim or survivor’s perspective, whether the image originated from a camera or from software, the impact is deeply traumatic and a violation of their right to be free from torture and inhuman or degrading treatment.”
Like with other child sexual abuse material, it is extremely difficult for children because of its permanence (endless copying, altering and redistribution) leaving victims with ‘no sense of control or closure’. Additionally, generative models can be trained and fine-tuned using abusive images which re-victimises survivors. The rise of AI-generated CSAM also poses a real risk to reporting systems and law enforcement capacity.
The IWF welcomes the UK Government’s action on enabling designated authorities like the IWF to test AI models, noting this is an important step in safety-by-design. However, they are calling for services (include AI providers) to take greater accountability to ensure their products are safe, with child safety ’embedded from the outset rather than treated as an afterthought’.
“We cannot tackle this risk alone. We need safety-by-design embedded as a non-negotiable standard in AI development with clear standards set by governments. This technology has extraordinary potential to change our lives for the better, but only if we mitigate the harms first. Injustice is not inevitable. We have an opportunity now. We must not waste it.” – Kerry Smith, CEO Internet Watch Foundation
Learn more
References
- IWF (2026) ‘Harm without limits: AI child sexual abuse material through the eyes of our Analysts’. Available at: https://www.iwf.org.uk/media/hl1nvdti/iwf-ai-csam-report-2026.pdf (Accessed: 27 April 2026)
- IWF (2026) ‘Key Findings: AI CSAM Statistics for 2025-2026’ Available at: https://www.iwf.org.uk/about-us/why-we-exist/our-research/how-ai-is-being-abused-to-create-child-sexual-abuse-imagery/ (Accessed: 27 April 2026)
