Deepfakes are synthetic images, videos, or audio created by artificial intelligence — designed to look indistinguishable from real content. In the context of child safety, AI deepfakes are being used in two primary ways: to generate fake intimate or sexually explicit images of children from ordinary photos, and to create convincing fake identities for grooming.
The critical point parents must understand: your child does not need to have sent any real images for this to happen. AI tools available for free online can take a single clear photo — a school portrait, a TikTok video frame, a family photo posted on Facebook — and generate a fabricated explicit image. These images are then used to blackmail, humiliate, or coerce the child.
In 2025, NCMEC received over 7,200 reports involving AI-generated child sexual abuse material — up from 4,700 in 2024. The IWF's 2026 AI CSAM report found a 26,385% increase in AI-generated child sexual abuse videos in 2025 alone: 3,443 videos were assessed, up from just 13 the year before, with 65% classified as Category A — the most severe under UK law. A UNICEF, ECPAT and INTERPOL study across 11 countries found at least 1.2 million children disclosed having photos manipulated into sexually explicit deepfakes in the past year, representing roughly one child per classroom in some countries.
The Grok AI lawsuit (xAI, July 2026) shows the scale possible with unfiltered tools: the chatbot generated roughly 3 million sexualised images in just 11 days — including 23,000 depicting apparent minors — made from ordinary photos, without the CSAM safety filters used by other AI companies. In one case, a single user created over 7,000 CSAM images of a minor victim using Grok. In Malaysia, cases are being prosecuted under the Computer Crimes Act 1997, Section 507 of the Penal Code (criminal intimidation), and the Sexual Offences Against Children Act 2017 for AI-generated CSAM.
This threat also intersects directly with sextortion: criminals increasingly use deepfakes as their opening threat, removing the need to trick a child into sharing real images at all.