Disclosures, platforms & rules
The councillor deepfake decision: an AI label is not the whole moderation question
When a platform decision concerns a deepfake, ask which rule supports the outcome. The fact that media is synthetic and the question of whether its content violates a policy require separate explanations.
The case and decision dates
On September 17, the Oversight Board published a decision about a Facebook video posted in November 2025 that falsely portrayed a Scottish councillor making statements about immigration. The Board called for removal under the Hateful Conduct policy. It also discussed AI labeling and misinformation rules, with differences among members on parts of the analysis. Its recommendations should not be mistaken for policies already implemented by Meta.
Sources: Oversight Board: Board explanation of the councillor-video decision
Read the decision in layers
A headline about ‘removing an AI video’ can conceal the decisive issue. To explain the outcome accurately, map the decision to the particular post, the rule applied and the remedies requested. Keep the fabricated message out of your correction unless its exact wording is necessary to understand a specific point.
- Record the original post period and the later decision date separately. The publication of a decision does not make the underlying clip a new statement.
- Identify what is established about the impersonation and how the decision attributes that finding. Do not reproduce the synthetic dialogue as an ordinary quote from the councillor.
- Find the policy named as the basis for removal. Separate that basis from the Board’s discussion of other rules or labels.
- Mark recommendations, member disagreements and the operative outcome as different categories in your notes. Avoid converting a recommendation into an implemented platform feature.
- When reporting a similar post, describe its actual context and harm using the platform’s available report categories. The existence of this decision alone does not establish the outcome of another case.
Correct the attribution without replaying the attack
Link to the Board’s explanation and, where relevant, the person’s original public statement. State that the disputed words were falsely attributed, rather than restating them in a headline that can circulate without its correction. An image extracted from a video cannot authenticate its speech. isGenAI does not offer a video or voice verdict for this incident, and an AI label would not independently establish whether the underlying message complies with a platform policy.
Questions about this case
Does adding an AI label settle whether a post should remain online?
No. A label addresses disclosure. A platform may separately assess impersonation, hateful conduct, harassment or other policy issues in the content and its context.
Sources and dates
- Board explanation of the councillor-video 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.