In a recent company blog post dated Jan. 6, Meta announced plans to introduce new standards concerning AI-generated content on its social media platforms, including Facebook, Instagram, and Threads, in the coming months.
Under these new standards, content identified as AI-generated, either through metadata or intentional watermarking, will be visibly labeled as such. Moreover, users on Meta platforms will have the ability to flag unlabeled content suspected of being generated by artificial intelligence.
This approach echoes Meta's early content moderation practices, harking back to the era before AI-generated content, when the company, then known as Facebook, developed a user-facing system for reporting content that violated the platform's terms of service.
In 2024, Meta is leveraging its vast user base across social networks to empower users with tools to flag content once again, potentially tapping into one of the largest consumer crowd-sourcing forces worldwide.
This initiative also places responsibility on creators using Meta's platforms to label their own content as AI-generated whenever applicable, as failure to do so may result in penalties, according to the blog post.
Meta states that whenever its built-in tools are utilized to create AI-generated content, such content receives a watermark and label clearly indicating its origin. However, not all generative AI systems come with these embedded guardrails.
To address this, Meta is collaborating with other companies through consortium partnerships, including Google, OpenAI, Microsoft, Adobe, Midjourney, and Shutterstock, to develop methods for detecting invisible watermarks at scale. However, these methods currently only apply to AI-generated images, as they have not been extended to audio and video at the same scale.
As a result, Meta acknowledges its current inability to detect AI-generated audio and video content at scale, including deepfake technology, as stated in the blog post.