Council says Meta’s rule on deepfake is incoherent – 02/05/2024 – Tech

Meta’s Oversight Board concluded that a Facebook video that falsely suggests that US President Joe Biden is a pedophile does not violate the company’s current rules, while also finding the policy “incoherent” and focused very restricted in content generated by artificial intelligence.

The board, which is funded by Meta but administered independently, took up the case over the Biden video in October in response to a user complaint about an altered seven-second video of the president posted on the company’s main platform.

Monday’s decision is the first to criticize Meta’s “manipulated media” policy, which prohibits certain types of deceptively edited videos, amid growing concerns about the potential use of new AI technologies to influence this year’s US presidential elections.

The policy “lacks persuasive justification, is incoherent and confusing to users, and does not clearly specify the harms it is intended to prevent,” the board said.

The body suggested that Meta update the rule to cover audio and video content, regardless of the use of AI, and apply labels identifying it as manipulated.

The council stopped short of calling for the policy to apply to photographs, warning that this could make the policy very difficult to enforce.

Meta, which also owns Instagram and WhatsApp, informed the board during the review that it was planning to update the policy “to respond to the evolution of new and increasingly realistic AI,” according to the decision.

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

Altered versions of the same video began going viral in January 2023, the board said.

In its decision, the Oversight Board said Meta was right to leave the video up in accordance with its current policy, which prohibits deceptively altered content only if it is produced by AI or makes people appear to be saying words. that they never actually said.

The council said content not altered by AI “is prevalent and not necessarily less misleading” than content generated by AI tools.