Meta introduces updated AI deepfake guidelines: Focus on labeling over removals

by

in

– Meta will label a wider range of AI-generated content, including deepfakes, with a “Made with AI” badge
– The company will not remove manipulated media unless it violates other policies, opting instead to add informational labels and context
– Meta will work with independent fact-checkers to identify risks related to manipulated content and reduce the reach of false or altered content on its platforms

Meta has announced changes to its rules regarding AI-generated content and manipulated media following criticism from its Oversight Board. Starting next month, a “Made with AI” badge will be used to label deepfakes, with additional contextual information shown when content has been manipulated in ways that could deceive the public on important issues. This move aims to provide transparency and context rather than removing manipulated media.

The changes could result in more misleading content being labeled, especially with many elections occurring worldwide. However, Meta will only apply labels to deepfakes with industry standard AI indicators or when the uploader discloses it as AI-generated. Content that falls outside these boundaries may remain unlabeled on the platform.

Meta’s decision to shift towards labels rather than takedowns for AI-generated or manipulated media aims to address content moderation concerns and systemic risks while protecting free speech. The company plans to stop removing content solely based on its current manipulated video policy, giving users time to understand the AI disclosure process before the policy changes in July.

Following feedback from its Oversight Board, Meta is expanding its policies to cover a broader range of AI-generated and manipulated content. The company plans to add more prominent labels to content that poses a high risk of deceiving the public, allowing users to better assess and understand the context of such content. Additionally, Meta will now rely on a network of independent fact-checkers to help identify and address risks related to manipulated content on its platforms.

Source link