AI-generated and AI-altered CSAEM

AI changes the speed, scale and realism with which sexualized material can be created, but it does not erase human responsibility. The person selecting a target, supplying an image, writing a prompt, distributing the output or using it to threaten a child is making choices that can produce serious harm.

Different forms require different analysis

• Identifiable-child deepfake: a real child’s face, body or likeness is placed into fabricated sexual content.
• Nudification or image alteration: an ordinary image is transformed to appear nude or sexually explicit.
• Real-image-derived output: an AI system transforms or extends an existing sexual abuse image.
• Wholly synthetic depiction: the output does not correspond to one known child, although real images may still have influenced the system.
• AI-assisted coercion: fabricated content is used to groom, harass, impersonate or sextort a child.
• Unspecified origin: the available evidence does not establish whether the material is real, altered or synthetic.

The harm is real even when the image is fabricated

An identifiable child may experience humiliation, fear, bullying, stalking, school disruption and loss of control over their identity. The material can be mistaken for authentic content and repeatedly redistributed. A fabricated image may also be used to force a child to pay money, send real content or remain in contact with the person making the threat.

Canadian and Ontario context

The Criminal Code definition in section 163.1 includes a person who ‘is or is depicted as being’ under 18 in specified visual representations. Canadian child-protection guidance states that sexually explicit AI-altered depictions of minors can engage criminal law. Ontario’s Provincial ICE Strategy and the Canadian Centre for Child Protection have jointly warned about sexually explicit deepfakes affecting youth and schools. Because technology and case law continue to develop, public legal statements should be checked immediately before publication.

If a child is targeted

• Believe the child and make clear that fabricated content does not make them responsible.
• Preserve the account, URL, threat messages, dates and non-prohibited source image information; do not download or redistribute the sexualized output.
• Report to local police or Cybertip.ca and use NeedHelpNow.ca for practical support with image-related online harm.
• Notify the school or organization when needed for safety, anti-harassment measures and evidence preservation, while limiting unnecessary circulation and disclosure.
• Treat demands involving a fabricated image as sextortion and follow the same urgent response: do not pay or send additional content.

What investigators and institutions must distinguish

Reports and public statements should identify whether the material is wholly synthetic, altered from an identifiable child, derived from real CSAEM or of unknown origin. They should also preserve information about the model or service, version, prompts, input images, account history and distribution pathway where that information can be obtained lawfully.