A Deepfake Put a Woman in Her Own Bedroom Selling Pills She Never Endorsed
What happened
In March 2024, Michel Janse, a Christian social media influencer with a following built around content on travel, home decor, and wedding planning, discovered a YouTube advertisement using her face without her knowledge or consent. The ad placed her likeness in what appeared to be her own bedroom, dressed in her own clothes, and used that setting to sell erectile dysfunction pills. She had not authorized the advertisement and had no prior warning it existed.
Experts who examined the advertisement speculated that the video had been generated by an AI system trained directly on Janse's existing posts. Her public content, accumulated over time as part of her normal creative output, had apparently become training data for a model capable of reproducing her face and manner convincingly enough to pass as genuine. The person who built the advertisement did not need footage she had sold or licensed. They needed only what she had freely published for her own audience.
This is the quality that separates AI-powered likeness theft from older forms of impersonation. Fabricating a convincing commercial featuring a specific person once required either direct access to footage under their control or significant production resources. A deepfake system trained on publicly available video collapses that barrier entirely. The person's own creative work becomes the raw input for a commercial product she had no role in approving and no means of anticipating.
Janse reported the advertisement to YouTube, which removed it. That sequence, a complaint followed by a platform takedown, is now the standard resolution path for these cases. It places the burden of detection entirely on the person whose face was used, requiring her to find the advertisement before she can challenge it, and it offers no visibility into whether the same underlying asset was distributed elsewhere or circulated further before the complaint landed.
The deeper problem sits upstream of any moderation queue. By the time Janse found the advertisement, the deepfake had already been built, deployed, and viewed. No process required the advertiser to demonstrate consent before the upload went live. No log recorded which model generated the video, what data trained it, or who commissioned the work. There is no provable record of what a system did, and without one, the person whose face was taken has no evidence chain adequate for legal action and no way to trace how widely the asset spread before the platform removed it.
Reported impact
- Affected parties
- Not publicly disclosed
- Harm type
- Not publicly disclosed
- Scale
- Not publicly disclosed
- Financial impact
- Not publicly disclosed
- Regulatory action
- Not publicly disclosed
Classification
Relevant governance controls
Governance control mapping is not available for this record.
- No controls mapped
Not publicly disclosed
Control mapping is analytical. It does not state that any control would have prevented the incident.
Sources and evidence
This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.