On Friday, Meta unveiled new policy revisions aimed at ceasing the censorship of harmless AI-generated content and instead opting to categorize a broader spectrum of video, audio, and image materials as ‘Made with AI.’
The decision to update Meta’s policies followed a scenario where a contentious post, depicting President Joe Biden in a controversial manner, was not removed despite being edited to imply inappropriate behavior. The post labeled Biden as a “pedophile,” prompting the Oversight Board to support Meta’s choice to retain the post while highlighting the shortcomings of Meta’s existing manipulated media policy as being too restrictive, inconsistent, and perplexing for users.
Previously, Meta’s stance focused on eliminating “videos that are fabricated or modified by AI to depict individuals uttering fabricated statements.” The Oversight Board cautioned that this approach failed to address other forms of manipulated media, such as “cheap fakes,” altered audio, or content portraying individuals engaging in fictitious actions.
Monika Bickert, Meta’s vice president of content policy, acknowledged the need for a broader perspective, stating, “It’s equally important to address manipulation that shows a person doing something they didn’t do,” aligning with the Oversight Board’s viewpoint.
Commencing in May 2024, Meta will affix “Made with AI” labels to any content identified as AI-generated and those self-disclosed as such by users.
Moreover, Meta will refrain from censoring content that does not breach community standards, opting for a more lenient approach by incorporating labels to provide additional context rather than removal.
The revised approach aims to offer users enhanced transparency and comprehension of the content they encounter, ensuring they can make informed assessments and recognize similar content across various platforms.
In July, Meta will halt the censorship of AI-generated content that does not infringe upon regulations concerning voter interference, harassment, violence, or incitement.
By embracing the Oversight Board’s counsel to establish a unified Manipulated Media policy, Meta intends to prevent a range of harms beyond mere misinformation, including safeguarding democratic processes like voting rights and public engagement.
The Oversight Board expressed satisfaction with Meta’s decision to expand the labeling of AI-generated content, emphasizing the importance of context and transparency while also removing posts that violate platform guidelines.
Looking ahead to the 2024 elections, the Oversight Board stressed the urgency of implementing these changes promptly to counter the dissemination of deceptive AI content that could impact electoral processes worldwide.
Meta’s consultation with diverse stakeholders and extensive public opinion research underscored broad support for labeling AI-generated content and limiting removal to instances of significant risk, acknowledging the increasing prevalence of generative AI in creative endeavors.
Despite challenges in detecting AI-generated content due to variations in industry signals and the potential removal of identifying markers by malicious actors, Meta remains committed to enhancing detection methods and enforcing penalties for non-disclosure by users.
As AI technology evolves and the proliferation of misleading content escalates, Meta and other platforms must adapt swiftly to maintain integrity and transparency in the digital landscape, as emphasized by the vigilant oversight of the Board.
