Superpower Daily

Superpower Daily

Your reading space

Sign in
Databricks Introduces Omnigent Beta for Agents That Run Across...OpenAI Says an Astra Training Model Inserted Its Own Jailbreak...Meta’s Oversight Board Orders Facebook to Remove Two AI Deepfak...European Commission Proposes Barring Chatbots From Simulating R...Moonshot Adds Professional Data Feeds to Kimi for Financial Ser...McKinsey and SEMI Forecast U.S. Chip Shortfall of Up to 157,000...Databricks Introduces Omnigent Beta for Agents That Run Across...OpenAI Says an Astra Training Model Inserted Its Own Jailbreak...Meta’s Oversight Board Orders Facebook to Remove Two AI Deepfak...European Commission Proposes Barring Chatbots From Simulating R...Moonshot Adds Professional Data Feeds to Kimi for Financial Ser...McKinsey and SEMI Forecast U.S. Chip Shortfall of Up to 157,000...
Swipe

Meta’s Oversight Board Orders Facebook to Remove Two AI Deepfakes

The binding removals concern a Scottish councillor and a Muslim campaign volunteer. The Board also wants labels to limit how deceptive AI posts spread.

By 3 min read
Meta’s Oversight Board Orders Facebook to Remove Two AI Deepfakes
Meta’s Oversight Board Orders Facebook to Remove Two AI Deepfakes

Listen to this story

The audio brief

About 1:32
0:001:32
Read transcript
Facebook must remove two AI-generated deepfake videos after Meta’s Oversight Board found that both violated the company’s rules—and called its safeguards for this kind of content “consistently and fundamentally inadequate.” One video falsely portrayed a Scottish Labour councillor making inflammatory claims about refugees. The Board said it alleged criminal and sexually predatory behavior by refugees as a group, violating Meta’s hateful-conduct policy. It also should have carried a high-risk AI label. Facebook had cleared the video and declined to label it, despite the Board bringing the case directly to the company. The Board pointed to audio that did not fully match the councillor’s facial movements as evidence it was AI-generated. The second video mocked a Muslim campaign volunteer involved in menstrual-health education in Europe. It fabricated scenes of her doing absurd exercises and eating junk food. The Board found that it breached bullying-and-harassment rules. Manipulated material targeting her accumulated tens of millions of views across platforms, although that figure covers more than the specific video being removed. The removals are binding because the Oversight Board is quasi-independent and its decisions must be followed. But its nine wider recommendations are not. They include warning screens, tougher penalties for repeat offenders, broader protection for private people falsely depicted, and—most significantly—demoting high-risk AI posts in feeds. The key question is whether Meta turns those recommendations into real limits on distribution.

Story brief

3 key points

Meta’s Oversight Board is forcing Facebook to take down two AI-manipulated videos after finding failures in both enforcement and labeling. One falsely depicted a Scottish Labour councillor making hateful claims about refugees; the other mocked a Muslim menstrual-health campaign volunteer. The removals are binding, but the Board’s nine broader recommendations—including feed demotion for high-risk AI content, warning...

  1. 01

    The Board called Meta’s deepfake safeguards “consistently and fundamentally inadequate.”

  2. 02

    One video should have been removed and labeled under Meta’s existing rules, but Facebook cleared it.

  3. 03

    The second case targets deepfakes portraying private people as saying or doing things they never did.

Meta’s Oversight Board has ordered Facebook to remove two AI-generated deepfake videos, concluding that both violated company rules. In the same rulings, the Board said Meta’s safeguards for addressing AI deepfakes were “consistently and fundamentally inadequate.”

One case involved a Scottish Labour councillor falsely portrayed as making inflammatory comments about refugees. The Board found that the video breached Meta’s hateful-conduct rules because it alleged criminal and predatory sexual behavior by refugees as a group. It also said the post should have received a high-risk AI label.

Facebook had concluded that the councillor video did not violate its policies and did not merit an AI label, even after the Board raised the case directly with the company. The Board said the clip appeared to be AI-generated because its audio was not fully synchronized with the councillor’s facial movements.

A second target, and a wider reach problem

The other removal concerns a video mocking a Muslim campaign volunteer who worked on menstrual-health education in Europe. It fabricated scenes of her doing absurd exercises and eating junk food, and the Board found it violated Meta’s bullying-and-harassment policy. AI-manipulated material targeting her drew tens of millions of views online, including on Meta platforms, though that total covers more than the individual video at issue.

The proposed shift: labels that affect reach

The Board recommended nine policy changes. Its main proposal would make a high-risk AI label alter distribution: Facebook would demote labeled posts in feeds, making them less likely to appear to users. It also recommended greater transparency about when Meta applies AI labels.

Other recommendations include

  • Broader circumstances for applying high-risk AI labels.
  • Warning screens that require a click before viewing AI-generated content.
  • Stronger penalties for accounts that repeatedly share deceptive AI content.
  • A wider definition of unwanted manipulated imagery for private people falsely depicted saying or doing something.

The two cases expose different gaps in Meta’s approach. The councillor video should have been removed and labeled under existing policy, the Board said. For the volunteer’s case, it wants Meta to explicitly cover deepfakes that make private people appear to say or do things they never did.

Editorial analysis

Our Read

The important divide in these rulings is between what Meta must do now and what it may choose to do next. The two removals are binding; the wider package is not. The Board’s recommendations would make a label consequential by connecting it to distribution, warnings and repeat-sharing penalties. That approach leaves some material online while reducing its reach, rather than treating moderation as a choice only between removal and no action. The next meaningful signal will be whether Meta adopts those recommendations, especially feed demotion and clearer rules for impersonated private people.

Related coverageMeta Ran Ads for a Tool Promoting Sexualized Political DeepfakesSan Francisco Sends Meta Cease-and-Desist Over AI Child Abuse Ads

Sources

  1. theguardian.comMeta ordered to remove UK deepfakes as oversight board criticises ‘inadequate’ safeguards

Loading discussion...

Meta’s Oversight Board Orders Facebook to Remove Two AI Deepfakes | Superpower Daily