AI-Generated Non-Consensual Intimate Imagery — Canadian AI Incident Monitor

Escalating Confidence: high Potential severity: Severe Version 1

A major AI platform generated over 3 million non-consensual sexualized images — including of minors — before safety controls intervened. Canada's Privacy Commissioner has expanded its investigation into X, but the legal framework for AI-generated NCII has significant gaps. This is targeted, personalized harm at industrial scale, with disproportionate impact on women and girls.

Identified: July 28, 2025 Last assessed: March 8, 2026

Description

Generative AI has made it possible to create realistic non-consensual sexualized imagery of any person from a single clothed photograph. The most dramatic demonstration occurred when xAI’s Grok chatbot generated approximately 6,700 “undressed” images per hour — over 3 million total — before the capability was restricted. Approximately 2% of those images depicted minors, crossing into child sexual abuse material territory.

The Privacy Commissioner of Canada expanded its ongoing investigation into X Corp in January 2026 to specifically address AI-generated sexualized deepfakes. The Commissioner’s testimony to the ETHI Committee highlighted AI-generated NCII as a priority concern.

The legal framework has significant gaps. Criminal Code section 162.1, which addresses non-consensual distribution of intimate images, was drafted before AI generation existed. Proving that an AI-generated image depicts an identifiable real person creates evidentiary challenges. No Canadian law requires AI platforms to prevent their systems from generating NCII, to test models against NCII generation capability before deployment, or to report when NCII generation occurs at scale.

The harm is gendered: research consistently shows that non-consensual intimate imagery disproportionately targets women and girls. The CEST documented in a 2024 report that deepfakes overwhelmingly target women, often in the form of non-consensual pornographic content. When AI makes this harm scalable and accessible, the impact on women’s participation in public life — political, professional, social — becomes a structural equality concern.

Risk Pathway

Generative AI enables the creation of non-consensual sexualized imagery of real individuals at unprecedented scale and accessibility. Unlike traditional image manipulation, current AI tools can generate realistic nude or sexualized images from a single clothed photo. The Grok chatbot generated approximately 6,700 "undressed" images per hour, with over 3 million generated before the feature was restricted. Approximately 2% depicted minors, crossing into CSAM territory. Canada's Criminal Code addresses some forms of non-consensual intimate images (s. 162.1) but was drafted before AI generation capabilities existed. The law requires proof that the image depicts a real identifiable person, creating evidentiary challenges for AI-generated content. No Canadian law requires AI platforms to prevent their systems from generating NCII, and no mandatory reporting obligation exists when such generation occurs at scale.

Assessment History

Mar 8, 2026 Escalating Confidence: high Severe

Grok generated over 3 million non-consensual sexualized images at a rate of ~6,700 per hour, with approximately 2% depicting minors. The OPC expanded its investigation into X to cover AI-generated sexualized deepfakes. Multiple "undressing" tools remain available on unregulated platforms. Canadian law has significant gaps — Criminal Code s. 162.1 was not drafted for AI-generated content. The ETHI Committee received testimony from the Privacy Commissioner on AI-generated NCII. Status is escalating because NCII generation tools are proliferating while governance remains inadequate.

Initial assessment. Status escalating based on confirmed industrial-scale NCII generation and proliferating tools.

Triggers

AI image generation models with safety filters removable or absent
"Undressing" tools becoming more accessible on unregulated platforms
No legal requirement for pre-deployment safety testing against NCII generation
Social media platforms hosting AI-generated NCII without detection

Mitigating Factors

OPC investigation creating regulatory scrutiny
xAI restricting Grok's NCII generation capability after public backlash
ETHI Committee study on AI examining NCII
Growing international regulatory attention (EU, UK, Australia)

Risk Controls

Explicit prohibition on AI platforms generating non-consensual sexualized imagery of identifiable individuals
Mandatory safety testing for image generation models against NCII generation
Criminal Code amendments addressing AI-generated NCII with provisions adapted for synthetic content
Platform liability for failing to prevent NCII generation at scale
Reporting obligations when AI systems generate NCII
Recourse mechanisms for victims of AI-generated NCII including expedited takedown