A Crucial Test Case for Facebook’s Deepfake Policies with a Fake Biden Video

Credit: Unsplash

Meta’s Oversight Board to Review Handling of Manipulated Biden Video

Facebook’s refusal to remove a leaked video of President Joe Biden will be investigated by Meta’s Oversight Board, an independent committee that monitors platform moderation. A major development. The video initially depicted President Biden during the 2022 midterms. It was modified to appear like he was repeatedly caressing a youngster, and the explanation was ambiguous. Meta kept the video.

The Oversight Board’s investigation is aimed at persuading Meta to create clear and robust regulations for dealing with biased media and electoral disinformation in time for the 2024 US presidential election and more than 50 other elections worldwide. Dan Chaison, the Board’s spokesman, underlined the need of election fairness and investigating how Meta might better resist changing material that tricks consumers, especially as AI improves.

Meta said in a blog post that the altered video did not violate their hate speech, abuse, or changed media guidelines. The Meta policy on manipulated media removes content that “has been edited or synthesized in ways that are not apparent to the average person and would likely mislead the average person to believe that a subject of the video said words that they did not say.” Meta stated that the movie footage was not altered using AI or machine learning.

Generative AI, which can create convincing false audio, video, and pictures, might make the 2024 elections more hazardous and challenging, say experts. Meta has claimed to fix creative AI’s issues, but watermarking hasn’t worked. Recent occurrences, like a Slovakian Facebook audio clip, illustrated how Meta’s biased media restrictions may disseminate bogus audio.

The Biden video wasn’t generated or edited by AI, but the Oversight Board wants public input, especially on AI’s participation. A bigger inquiry of Meta’s revised video policy is underway.

“A fake Biden video shows the limits of Meta's deepfake policies”https://t.co/imCCtbXL9G

— Ralf Flueggen (@flueggen) October 10, 2023

Sam Gregory, program director at Witness, underlined the need for guidelines to address mis-contextualized and mis-edited media in shared content.

The Oversight Board’s recommendations are binding on Meta. This research is supposed to develop Meta’s 2024 election strategy, according to previous assessments of its global election efforts.

Some worry that the Oversight Board will influence how Meta uses its money and resources on these issues, especially in foreign elections. Meta, like other digital firms, has been attacked for reducing trust and safety staff, which handles bogus news and hate speech. The corporation has also struggled to monitor non-English or US-centric content.

Google’s features help consumers recognize AI-created or manipulated photos. However, Meta has not built any tools for users or fact-checkers to learn more about content.