Matériel d'exploitation sexuelle d'enfants généré par l'IA au Canada

Generative AI is enabling the production of child sexual abuse material at a scale and speed that outpaces existing detection and enforcement infrastructure. The Canadian Centre for Child Protection has documented increasing volumes of AI-generated CSAM. Existing hash-based detection systems like PhotoDNA — designed to identify known images through digital fingerprints — cannot detect AI-generated content because each synthetic image is unique.

The legal framework presents additional challenges. Canada's Criminal Code provisions on child pornography were drafted for human-produced material. While the provisions may apply to purely synthetic AI-generated CSAM depicting no identifiable real child, prosecution is still in early stages — Steven Larouche of Sherbrooke, Quebec was sentenced in 2023 to over three years for creating deepfake child pornography, in what the presiding judge described as the first such case in Canada. The gap between generation capability and detection capability is widening: producing realistic synthetic CSAM requires minimal technical expertise and no access to real children, while detecting it requires investments in new technology that law enforcement agencies have not yet made.

This hazard is the most acute current manifestation of a broader structural pattern: generative AI content production capability outpacing institutional detection and response capacity. The asymmetry between cheap, scalable generation and expensive, fragile detection applies across harm categories, but the consequences are most severe when the content involves child exploitation.

AI developers have implemented content policies prohibiting CSAM generation, and some platforms have added technical safeguards to prevent their models from producing such content. International efforts to develop detection tools for AI-generated imagery are underway. The challenge lies in the gap between platform-level controls and the availability of open-source models that lack equivalent safeguards.