Facebook Collaborates with Industry Partners to Identify and Label AI-Generated Content
Facebook is taking steps to increase transparency and provide users with the ability to identify AI-generated content on its platform. The company is currently developing technical standards in partnership with other industry giants, including Google, OpenAI, Microsoft, Adobe, Midjourney, and Shutterstock, to mark and identify images created using artificial intelligence (AI) technologies.
To indicate when images are AI-generated, Facebook is exploring different methods such as visible markers, invisible watermarks, and embedded metadata. These markers will provide users with insights into whether the images they see are the result of AI algorithms. While progress has been made in identifying AI-generated images, the detection of AI-generated audio and video remains a challenge.
To address this, Facebook plans to introduce a feature that allows users to disclose and label AI-generated video and audio content. Organic content with realistic video or audio that has been digitally created or altered will be required to be properly labeled. By doing so, Facebook aims to maintain transparency and ensure users are able to distinguish between AI-generated and non-AI-generated content.
Furthermore, Facebook acknowledges that as the spread of AI-generated content increases, adversarial techniques may emerge. As a result, the company is exploring methods to automatically detect AI-generated content and make it more difficult to remove or alter invisible watermarks.
To help users identify AI-generated content, Facebook advises them to consider factors like the trustworthiness of the accounts sharing the content and unnatural details that may suggest the use of AI algorithms. Additionally, Facebook emphasizes that its Community Standards apply to all content, whether AI-generated or not, and artificial intelligence plays a crucial role in enforcing these standards.
In line with this, Facebook is currently testing Large Language Models (LLMs) trained on its Community Standards to more accurately and effectively identify policy violations. Furthermore, AI-generated content can be fact-checked by independent fact-checking partners, and any debunked content will be properly labeled.
Facebook is committed to developing generative AI tools responsibly and transparently. The company values user feedback and continuously collaborates with industry partners such as the Partnership on AI (PAI) to ensure that AI technologies are utilized in a manner that benefits and respects its users.
