Meta, the parent company of Facebook, has been directed by its independent oversight body to remove deepfake videos targeting a British politician and a young Muslim activist from its platform. This move underscores the growing concern surrounding AI-generated fake content and its impact on public discourse.
The oversight board identified a contentious video featuring a Scottish Labour party councillor as particularly problematic. The clip portrayed the councillor making inflammatory statements about refugees, which the board deemed unacceptable. Additionally, an AI-generated video misrepresenting a Muslim campaign volunteer—depicting her in absurd scenarios while offering health advice—was also flagged for removal.
The board criticized Meta’s current safeguards as “consistently and fundamentally inadequate” in addressing the alarming rise of deepfakes. Pamela San Martin, co-chair of the oversight board, noted, “AI-generated deepfakes are increasingly being used to harass and silence women from engaging in public discourse.”
In the flagged Scotland video, the councillor was incorrectly represented as making egregious claims about refugees. The board suggested that the video’s audio was not synced with the councillor’s facial movements, indicating it may have been artificially produced.
Despite the board’s inquiries, Meta initially rejected claims that the video violated its content policies or merited an AI warning. However, the oversight board later asserted that the video contravened Meta’s guidelines on hate speech by generalizing criminal behavior associated with an entire group.
The board called for a “high-risk AI” label for such content and advocated for the company to strengthen its approach to deepfake management. Their findings indicated a need for enhanced policies, including more clear definitions of “high risk,” harsher penalties for repeat offenders, and greater transparency in marking AI-generated content.
Recommended changes from the board include using algorithms to prioritize the demotion of “high risk” content, limiting user exposure to AI-generated deepfakes through additional warning mechanisms, and revising policy definitions to encompass manipulated imagery of private individuals.
In a related decision, the board mandated the removal of another AI-generated video ridiculing a Muslim woman engaged in promoting menstrual health awareness. This content had garnered millions of views and was deemed a violation of Meta’s bullying and harassment policies. The board recommended a shift in policy to better address instances of deepfakes targeting individuals.
The oversight board was established by Meta in 2020 to act as an adjudicator for content issues on its platforms. The company is yet to respond to inquiries regarding these developments.
Editor’s Take
This ruling highlights a critical stand against the misuse of AI technologies in spreading misinformation and harassment, particularly targeting vulnerable individuals. As social media platforms grapple with the ethical implications of AI-generated content, this could catalyze more robust regulatory frameworks and inspire similar actions across the industry, ultimately fostering a safer online environment for all users.
Source: www.theguardian.com
