Meta is under renewed pressure to clarify how it moderates AI-generated explicit images, after its Oversight Board examined two cases involving public figures on Instagram and Facebook.
The Board’s message is direct: the company’s current rules are too narrow, too dependent on older language, and too easy for serious reports to fall through. Its recommendations focus on how Meta names the harm, where it places the policy, and how quickly it can detect repeat uploads.
Why the Oversight Board wants Meta to change its wording
At the center of the Board’s recommendations is language. Meta’s current approach to AI-generated explicit images is tied to a rule about “derogatory sexualized Photoshop” in the Bullying and Harassment section of its policies.
The Board wants Meta to replace “derogatory” with “nonconsensual.” That change matters because the central issue is not whether an image insults someone. It is whether a sexualized image was created, altered, or shared without consent.
The Board also urged Meta to stop using the word “Photoshop” as the policy term. The concern is that the word points to one familiar form of image editing while the problem now includes broader forms of manipulated media, including AI-generated content.
In addition, the Board recommended moving these rules from the Bullying and Harassment section to the “Sexual Exploitation Community Standards” section. That would place AI-generated explicit images closer to policies that directly address sexual abuse and exploitation, rather than treating them primarily as harassment.
The policy gap around private and non-commercial images
Meta currently prohibits nonconsensual imagery if it is “non-commercial or produced in a private setting.” The Oversight Board said that condition should not be required for Meta to remove or ban AI-generated or otherwise manipulated explicit images created without consent.
The Board’s concern is practical. If an image is synthetic or manipulated, the harm does not depend only on whether it was produced in a private setting or whether it was commercial. The core issue is the lack of consent.
That distinction is especially important for AI-generated explicit images because they may not come from an original private photo. A person can still be depicted in an explicit way without agreement, and the resulting image can still spread across platforms.
Two cases showed different outcomes inside Meta’s systems
The recommendations followed two high-profile cases involving explicit AI-generated images of public figures.
One case involved an AI-generated nude image of an Indian public figure posted on Instagram. Several users reported it, but Meta did not remove it. The company closed the ticket within 48 hours without further review. Users appealed, and the ticket was closed again.
Meta acted only after the Oversight Board took up the case. The company then removed the content and banned the account.
The second case involved an AI-generated image resembling a public figure from the U.S. that appeared on Facebook. In that instance, Meta had already placed the image in its Media Matching Service, known as MMS. The repository stores images that violate Meta’s terms of service and can help detect similar uploads.
Because that image was already in MMS due to media reports, Meta removed it quickly when another user uploaded it to Facebook.
The contrast between the two cases is important. The image of the Indian public figure was added to MMS only after the Oversight Board pushed Meta on the matter. Meta told the Board the image had not been in the repository earlier because there were no media reports about it.
Why relying on visibility can leave victims exposed
The Oversight Board warned that this approach creates an uneven system. Public attention can help a harmful image enter Meta’s detection tools, but people who are not widely covered by media may not get the same protection.
“This is worrying because many victims of deepfake intimate images are not in the public eye and are either forced to accept the spread of their non-consensual depictions or report every instance,” the Board said in its note.
That risk becomes larger when moderation depends heavily on user reports. If each new upload must be reported again, the burden shifts to the affected person or to other users who notice the content.
Devika Malik, a platform policy expert who previously worked in Meta’s South Asia policy team, told TechCrunch earlier this year that platforms largely rely on user reporting for taking down nonconsensual imagery. She said this may not be reliable for AI-generated media.
Malik also said the process can place an unfair burden on the affected user to prove identity and lack of consent. With synthetic media, she noted, that process can become more error-prone while harmful content gains traction.
Experts want clearer reporting and more context
Breakthrough Trust, an Indian organization that campaigns to reduce online gender-based violence, told the Oversight Board that Meta’s policy choices have cultural implications. The organization said nonconsensual imagery is often treated too lightly as an identity theft issue rather than as gender-based violence.
Barsha Chakraborty, the head of media at Breakthrough Trust, wrote to the Board that victims can face secondary victimization while reporting cases in police stations or courts. She also warned that once an image is online, removing it from the source platform may not be enough because it can spread quickly elsewhere.
Chakraborty told TechCrunch that users often do not know when their reports have been automatically marked as “resolved” in 48 hours. She said Meta should not apply the same timeline to every case and should build more user awareness around these issues.
Aparajita Bharti, founding partner of Delhi-based think tank The Quantum Hub (TQH), said Meta should let users provide more context when reporting content. Her point was that users may not understand how Meta divides rule violations across different policy categories.
Bharti said systems should prevent real issues from falling through the cracks because of technicalities in Meta content moderation policies.
Meta said it will review the Oversight Board’s recommendations. The next question is whether the company will treat AI-generated explicit images as a narrow moderation category or as a broader nonconsensual imagery problem that requires clearer reporting, faster matching, and less dependence on public visibility.