What has changed in the AI-CSAM landscape? IWF AI-CSAM Report Update July 2024
The Internet Watch Foundation’s July 2024 Report Update (the “Report”) highlights the steep and continued growth in the production and circulation of AI-Child Sexual Abuse Material (“AI-CSAM”). For children accused of sharing AI-CSAM, the blurring of boundaries between synthetic AI-CSAM and real abuse imagery creates new legal risks.
Details
The IWF recorded a 380% year-on-year increase in actionable AI-CSAM reports between 2023 and 2024[1]. In its snapshot study of a dark web AI-CSAM forum over a 30-day period (from 9 March – 7 April 2024), the IWF found that 90% of the images reviewed were realistic enough to be treated as pseudo-photographs under Section 1 Protection of Children Act 1978[2] (the “PCA”). Of those images, 32% depicted activities that would be classified as Category A according to the Sentencing Council’s Sexual Offences Definitive Guideline[3], the highest level of offender culpability.
- Open-source text-to-image models remain readily accessible and, because they can be used offline, make it possible for children to generate large volumes of AI-CSAM without detection or moderation.
- With all realistic AI-CSAM images being treated as indecent pseudo photographs under the PCA, and all unrealistic AI-CSAM images treated as prohibited images under Sections 62-68 of the Coroners and Justice Act 2009[4], children accused of AI-CSAM offences may be charged for conduct not originally contemplated by statute – especially as AI tools keep developing.
- Moreover, guidance exchanged online regarding the creation of AI-CSAM is not currently criminalised in England and Wales. The Section 69 Serious Crime Act 2015[5] offence of possessing a paedophile manual is limited to advice about real children, and the previous Government’s commitment to extend this offence to AI has not been passed into law[6].
- The IWF accordingly reiterates its recommendations from its October 2023 AI-CSAM Report[7]: criminalise manuals explaining how to create AI-CSAM, outlaw training AI on Child Sexual Abuse Material and mandate “child-safety-by-design” testing principles before releasing powerful AI models.
Commentary
Children accused of generating or sharing AI-CSAM are increasingly at risk of criminalisation for behaviour not contemplated under the existing legislative framework. Children can now download an open-source AI model, follow an online tutorial, and produce hundreds of Category A AI-CSAM pseudo-photographs in a single session – often without grasping the legal consequences.
Practitioners who work with children accused of creating AI-CSAM should:
- Scrutinise charging decisions: ensure distinctions are made between pseudo-photographs and prohibited images, for example.
- Consider mitigation: the Sentencing Children and Young People Definitive Guideline[8] requires an individualistic approach
- Explore diversion: out-of-court disposals are likely to be more in line with the principles of a Child First youth justice system to avoid unnecessary criminalisation and treat children as children first and foremost.
- Keep updated on imminent statutory change: for example, the incoming Crime and Policing Bill[9], if passed, is expected to create a specific offence of generating sexual deepfakes[10].
AI has transformed the landscape of child sexual abuse imagery offences, exposing children to new forms of criminal liability. Until the legislative framework is modernised, practitioners working with children accused of creating AI-CSAM should be mindful of the above tips, the commitment of the actors in the youth justice system to a Child First approach – one that continues to see children as children – and that recognises their welfare.
[2]Protection of Children Act 1978
[3]Sexual Offences Definitive Guideline, The Sentencing Council
[4]Coroners and Justice Act 2009
[6]Ministry of Justice, Laura Farris – Government cracks down on ‘deepfakes’ creation (16 April 2024)
[7]How AI is being abused to create sexual abuse imagery? IWF October 2023