Disclosures, platforms & rules
AI harassment of a private person: lessons from the September Board decision
A manipulated interview can falsify a person’s words and actions while leaving enough genuine footage to look familiar. Verification should restore attribution and context without giving the abusive edit a new audience.
A distinct harassment case
In a separate September 17 decision, the Oversight Board addressed AI-manipulated imagery of a private individual who had appeared in a television interview during a health campaign. The alteration fabricated words and actions to ridicule her. The Board called for removal under bullying and harassment rules and recommended stronger protections against unwanted AI imagery of people outside public life.
Sources: Oversight Board: Board explanation of the private-person harassment decision
Compare context without circulating the altered clip
The important comparison is between the original interview’s meaning and the claims attributed by the altered post. A person’s appearance is not evidence of what they said. Publicly naming the targeted individual or repeating the insulting portrayal may be unnecessary to explain the manipulation.
- Find a legitimate source for the original interview and record its date and subject. Do not assume that every frame in the altered version is fabricated.
- Describe the disputed changes in neutral terms: replaced speech, invented actions or a false role. Avoid reproducing ridicule or inferring personal traits from the image.
- Preserve the harmful post’s URL, timestamp and reporting reference. If documenting repeated posts, keep a private chronological list rather than a public compilation.
- Use the platform’s relevant harassment or impersonation reporting route and explain the difference from the source interview. A generic statement that the clip ‘looks AI’ is less informative than identifying the false attribution.
- Before sharing evidence with anyone beyond the reporting channel, consider whether identifying details are needed and whether the affected person wants them circulated. Retain only what is necessary for the specific report or appeal.
What a responsible public correction includes
State that an interview was manipulated, link to the authoritative decision and explain the verification method without embedding the abusive copy. Keep a policy recommendation distinct from a feature already available in the reporting interface. A detector result is not required to describe a demonstrated mismatch with the original interview, and isGenAI’s image tools cannot establish the authenticity of the video’s speech.
Questions about this case
Must I publish the altered video to explain that an interview was manipulated?
No. A neutral description of the mismatch, an authoritative source and a link to the appropriate original context can explain the issue without redistributing the abusive edit.
Sources and dates
- Board explanation of the private-person harassment decision — Oversight Board, 17 September 2026.
Sources checked on 27 September 2026. Linked reporting and official statements establish the attributed facts; the review steps are isGenAI’s analysis.
Our methodology explains supported checks and result limits. Found a factual error or a newer primary source? Use our corrections process and include this article’s URL.