Meta’s Oversight Board Isn’t Clear on Its Deepfake Rules: What It Means

NEW YORK: Meta’s oversight board found that a Facebook video falsely suggesting that US President Joe Biden was a pedophile did not violate the company’s existing rules, while calling them “incoherent” and too narrowly focused on AI-generated content .

The committee, which is funded by Meta but operates independently, took over the Biden video case in October in response to user complaints about an altered seven-second video of the president posted on Meta’s flagship social network.

Monday’s ruling is the first to target Meta’s “manipulated media” policy, which bans certain types of doctored videos, amid growing concerns that new artificial intelligence technology could influence this year’s election.

The committee said the policy “lacked a persuasive rationale, was incoherent, confused users and failed to clearly state the harms it sought to prevent”.

The board recommended that Meta update its rules to cover audio and video content, whether or not it uses artificial intelligence, and apply labels to identify it as being manipulated.

It stopped short of calling for the policy to apply to photos, warning that doing so might make the policy difficult to enforce at Meta’s scale.

Meta, which also owns Instagram and WhatsApp, told the board during the review that it planned to update its policy “to address the development of new and increasingly realistic artificial intelligence,” according to the ruling.

The company said in a statement on Monday that it was reviewing the ruling and would respond publicly within 60 days.

This video on Facebook doctored real footage of Biden exchanging “I Voted” stickers with his granddaughter and kissing her cheek during the 2022 US midterm elections.

The committee said that as early as January 2023, a modified version of the same video clip began going viral.

In its ruling, the oversight board said Meta was right to keep the video under its current policy, which only bans misleadingly altered videos if they were produced by artificial intelligence or make people appear to say Saying things they never actually said.

The committee said content modified by non-AI “is common but is not necessarily less misleading than content generated by AI tools.”

The policy should also apply to audio-only content and videos depicting people doing things they never actually did, the company said.

It added that enforcement should include applying labels to content, rather than Meta’s current approach of removing posts from its platform.

(This story has not been edited by News18 staff and is published from a Yonhap news agency feed – Reuters)