TikTok is ramping up its efforts to automatically label AI-generated content in its app, even when it was created with third-party tools. The company announced plans to support content credentials, a kind of that indicates the use of generative AI.
TikTok’s rules already require creators “realistic” AI-generated content. But that policy can be difficult for the company to enforce, particularly when creators use other companies’ AI tools. But because are increasingly used across the AI industry, TikTok’s new automated labels should be able to address some of those gaps.
Often described as a “nutrition label for digital content,” content credentials attach “tamper-evident metadata” that can trace the origins of an image and AI tools that were used to edit it along the way. That history can then be viewed by users if they come across a piece of AI-made content on a platform that supports the technology.
TikTok says it will be the first video platform to support content credentials, though it will take some time before these labels become commonplace since many companies are only just beginning to support the technology. (Google, Microsoft, OpenAI and Adobe have all pledged to support content credentials. Meta has said its using the standard to power labels on its platform as well.)
However, it’s worth noting that content credentials and other systems that rely on metadata aren’t foolproof. OpenAI notes on a that the tech “is not a silver bullet” and that metadata “can easily be removed either accidentally or intentionally.” Labels also simply aren’t that effective if people don’t bother reading them. TikTok says it has a plan to address that too. The company has partnered with fact-checking organization MediaWise and human rights organization Witness on a series of media literacy campaigns meant to educate TikTok users about the labels and “potentially misleading” AI-generated content.
Read the original article here