Consent at the Centre: Meta’s Oversight Board Decision Reflects WITNESS’ Key Concerns on AI-generated Sexualized Impersonation
The Meta Oversight Board, an independent body that reviews the company’s content moderation decisions, today overturned Meta’s decision to keep online an AI-generated sexualized video that had been reported by users, citing concerns about the enforcement of its policies on non-consensual intimate imagery (NCII). The ruling aligns with recommendations submitted by WITNESS in a public comment to the Board earlier this year.
The case involved an Instagram video depicting a woman that was reported as pornographic but only age-restricted by Meta, which determined that it did not violate the platform’s “Adult Nudity and Sexual Activity” policy. In its decision, the Oversight Board reversed Meta’s determination and raised broader concerns about the company’s handling of non-consensual intimate imagery and its reliance on user reports to identify and remove harmful content, reflecting points raised by WITNESS in its submission.
WITNESS Recommendations to the Board
WITNESS’ submission argued that Meta’s current approach is structurally insufficient because it fails to recognize AI-generated NCII as a form of technology-facilitated gender-based violence (TFGBV), relies too heavily on automated systems that under-prioritize harm, and overlooks the central issue of consent. The submission also highlighted the need for the Oversight Board to reiterate its 2024 recommendations, many of which remain unimplemented. WITNESS also emphasized the need to reframe cases such as the one in question in light of emerging examples of synthetic NCII and the lack of consent in the generation of this content.
The second part of our submission highlighted that AI-generated NCII has evolved into a scalable, commercialized, and mainstream form of TFGBV that disproportionately targets women, LGBTQ+ individuals, and human rights defenders. This abuse is no longer merely episodic but is fueled by systemic structural failures, including the proliferation and mass access to deepfake tools, the exploitation of likeness via smart devices, and the failure of platforms to differentiate between transformative expression and non-consensual sexual exploitation. Current moderation practices fall short because they rely on automated systems that under-prioritize harm, fail to recognize that the violation occurs at the moment of creation, and mistakenly rely on contextual cues like public figure status as a proxy for consent.
To address these vulnerabilities, platforms must fundamentally reframe their approach by recognizing AI-generated NCII as sexual exploitation and moving these prohibitions from “Bullying and Harassment” policies into “Adult Sexual Exploitation” standards. WITNESS recommends that platforms adopt a presumption of non-consent for realistic impersonations, mandate human review for all reports, and create dedicated reporting pathways rather than relying on ineffective age-gating or auto-closure mechanisms. Furthermore, platforms must implement product-level safeguards to prevent abusive generation at the system design stage and extend product accountability to AI-enabled capture devices to prevent covert likeness capture from fueling the abuse pipeline.
Based on these concerns, our recommendations to the Board were the following:
- Recognize AI-generated NCII as technology-facilitated gender-based violence, aligning policy language accordingly.
- Adopt a presumption of non-consent for realistic AI-generated sexualized impersonations of identifiable individuals, regardless of the fact that such images might have been generated based on media made available by the individuals themselves – as it implies a manipulation of the purposes regarding the image uses.
- Mandate human review for all AI sexual impersonation reports.
- Create a dedicated reporting pathway for AI-generated sexualized impersonation.
- Implement product-level safeguards to prevent the generation of sexualized impersonations at the system design stage.
- Extend product accountability to AI-enabled capture devices. Meta should establish responsible design standards for its hardware products to prevent covert likeness capture from feeding into exploitation pipelines.
Meta Oversight Board Decision
The Board determined that the post violated the company’s Adult Sexual Exploitation policy, concluding that AI-generated impersonation of real people should be considered non-consensual by default, regardless of whether the depicted individual reports the content themselves.
The investigation highlighted significant gaps in Meta’s current enforcement mechanisms, particularly concerning non-public figures who face disproportionate harm from such deepfakes. Meta had initially declined to remove the content, citing a lack of clear signals that the individual was a real person, as the depicted woman had not filed a report. The Board argued that this reliance on victim self-reporting places an undue burden on survivors, who may lack the resources or desire to engage directly with the platform’s reporting flows, especially in cases where the victim may no longer have an active account.
To address these enforcement gaps, the Board issued several key recommendations to Meta. They urged the company to explicitly add “AI-generated sexualized impersonation” as a valid signal for lack of consent under its Adult Sexual Exploitation policy. Furthermore, the Board recommended that Meta allow users to designate “connected accounts”—such as family or trusted associates—to report violations on their behalf. They also called for the implementation of a distinct reporting category for AI-generated sexualized impersonation, separate from generic “harassment” or “nudity” categories, to be made available to users globally.
The full set of recommendations is as follows:
Policy
1. To help ensure violating content is removed, Meta should add a new signal for lack of consent in the Adult Sexual Exploitation policy: context that content is AI-generated sexualized impersonation of real people. The Board will consider this recommendation implemented when both the public-facing and internal guidelines are updated to reflect this change.
Enforcement
2. To reduce the burden on victims of non-consensual intimate image abuse, Meta should allow users to designate connected accounts, such as those of trusted friends, family members or associates, that can report on their behalf potential Community Standards violations involving non-consensual intimate imagery, including impersonations. The Board will consider this recommendation implemented when Meta makes these features available and easily accessible to all users via their account settings.
3. To ensure effective and accessible reporting pathways to report on AI-generated non-consensual intimate imagery abuse, Meta should include AI-generated sexualized impersonation as a separate category in standard content reporting and appeal forms, distinct from “harassment” or “nudity.” These uniform reporting forms should be available globally. The Board will consider this recommendation implemented when Meta updates its reporting forms, shares data showing the global rollout results, and explains how the new reporting flows impact enforcement of the Adult Sexual Exploitation Community Standard.
This decision underscores the urgent need for systemic changes in how platforms manage the proliferation of AI-generated sexual abuse, which research shows overwhelmingly targets women and girls. As governments worldwide implement stricter regulations on deepfakes and AI-generated content, the Board emphasized that effective enforcement must account for local cultural nuances and the severe, long-term psychological and reputational harm caused by non-consensual imagery. Ultimately, the Board reiterated that robust safeguards and better content provenance tools are essential to protect users from the speed and sophistication of modern AI-driven exploitation.
Looking Ahead
We trust that this Oversight Board’s decision represents a crucial step forward in holding platforms accountable for the proliferation of AI-generated sexual abuse. By confirming that AI-generated impersonation should be treated as non-consensual by default, the Board validates the urgent need for a paradigm shift in content moderation. As Meta moves to implement these recommendations—including dedicated reporting pathways and better safeguards against synthetic abuse—it must ensure that its policies move beyond reactive measures. Protecting users requires a proactive, rights-respecting approach that acknowledges the severity of TFGBV and prioritizes the safety and dignity of all individuals over the speed of automated enforcement.