AI-Generated Child Abuse Material (AI CSAM): What schools need to know  

The Internet Watch Foundation (IWF) recently published their Artificial Intelligence Child Sexual Abuse Material (AI CSAM) report 2026. It gives us a great insight to this emerging safeguarding issue.

AI CSAM is the umbrella term for image or video content depicting the sexual abuse of children that has been created with assistance from, or entirely by, generative AI systems. While the proportion of AI CASM remains comparatively small among the huge amount of Child Sexual Abuse Material, it is becoming increasingly extreme with 65% of AI video content identified in 2025 classified as Class A (the most extreme classification).

The IWF report explains that AI CASM causes real harm as it fuels sexual interest in children, normalises extreme violence and increases the risk of contact sex offending. It is highly gendered, with 97% of illegal AI generated images showing girls being overwhelmingly targeted.

How is it generated?

The key AI models that are of concern to the IWF are:

Large language models (LLMs)

LLMs are AI systems trained on large text datasets to identify patterns and generate naturalistic text. Whilst LLMs don’t create images or videos but can be misused to describe harmful scenarios. These systems have dramatically risen in popularity and sophistication in the past few years, with popular examples including OpenAI’s ChatGPT and Google’s Gemini.

Nudify apps

Nudify tools use AI to generate nude images from clothed photos. Though marketed for adults, they’re increasingly misused to create indecent images of children, including by other children. There is no positive use case for these tools. They only serve to humiliate, harass and exploit and can serve to inflict further abuse.

The rise of generative AI tools has significantly increased the ease of producing realistic sexual deepfakes, with nude deepfakes making up around 98% of all deepfakes.

Text-to-image models

Text-to-image models have the ability to generate images from written text prompts, which can be misused to create realistic and harmful content. These tools can be misused to create realistic images very quickly. Users can refine prompts to make images more specific or extreme. Popular text-to-image models include Midjourney Inc’s Midjourney AI and Stability AI’s Stable Diffusion.

What does the Legislation say about this?

In February 2025, the UK Government announced measures under the Crime and Policing Bill to close legal loopholes related to AI-generated CSAM. This includes the introduction of a new criminal offence for making, adapting, possessing, supplying or offering to supply a CSA image-generator. At time of publication, the Crime and Policing Bill is at Report Stage in the House of Lords and is likely to receive Royal Assent in the coming weeks.

Where does AI CSAM appear?

AI CSAM is hosted and distributed across a range of online environments, including dark web forums and the clear web, demonstrating that this is a current and active threat.

How is AI CSAM harmful?

AI-generated sexual imagery of children causes real and lasting harm to victims, even if the physical abuse depicted did not take place. A defining characteristic of AI CSAM is that it actively widens the spectrum of harm by making it possible to fabricate scenarios. Perpetrators can design entirely novel forms of exploitative material tailored to their personal fantasies. 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.

There have been claims that AI CSAM causes less harm than ‘real’ CSAM, either because it does not directly involve offline offending against children or because entirely synthetic material is said to have no ‘real’ victim. However, IWF argue there is no evidence to support the claim that AI CSAM provides an effective harm-reduction method for offenders.

AI tools are often also used to manipulate existing child sexual abuse material involving survivors. Perpetrators have been known to take existing abuse images and use AI to generate new, frequently more extreme, versions. For survivors, this amounts to being abused again and again, years after the original harm, with no ability to stop it.

Research suggests that the consumption of realistic synthetic material reinforces existing sexual interests in children rather than providing a safe substitute and may increase the risk of escalation. Engagement with any form of CSAM has been shown to reinforce existing sexual interest in children.

There is a clear link between CSAM consumption, de-sensitisation and lower inhibition towards further online, or offline sexual offending.

What about chatbots?

The IWF also has concerns regarding the emergent use of chatbot technologies, including by young people and children.

Generative AI chatbots are software tools built on AI techniques that can simulate and respond to ‘human’ conversation. Even though chatbots like ChatGPT are designed with strict rules to block harmful or illegal use, people are still finding ways to use them indirectly to support that activity. People aren’t using chatbots to create illegal content directly—they’re using them to learn how to do it themselves.

Additionally, text-based chatbots alone carry significant risks, including holding the potential to provide offenders information on how to commit child sexual abuse and allowing them to simulate sexual conversations with children.

Conclusion

Worryingly, the IWF saw over 260 times more AI-generated child sexual abuse videos in 2025 compared to 2024 data.

For DSLs AI CSAM can no longer be viewed as a hypothetical change in the way children can be abused. These technologies create new risks that didn’t exist before and highly realistic, full-motion videos depicting child sexual abuse are now routinely identified online.

How can DSLs respond to these risks?

• Ensure AI CSAM is recognised within your school’s safeguarding framework by updating child protection and online safety policies and procedures

• Undertake ongoing training and updates regarding AI related safeguarding risks

• Ensure governors, school leaders and staff understand the issue and risks of AI CSAM

• Recognise that creating/sharing AI CSAM likely meets the ‘harm threshold’ and refer to children’s social care where appropriate

• Child-on-child abuse must be taken seriously, so treat any AI-generated sexual images of peers as abuse

• Work with IT Leads to ensure the school has appropriate filtering and monitoring systems

• Maintain accurate safeguarding records of all concerns (including AI CSAM)

• Ensure online safety and AI risks are embedded in PSHE lessons

About the author: Sophie Baker is a senior social work leader with over 20 years experience in frontline and practice development roles.