TikTok, the popular video-sharing platform, has announced its plans to deploy a new digital watermark called Content Credentials. This watermark will be used to identify images and videos created with the assistance of artificial intelligence (AI) that are submitted to the platform. While this move by TikTok aims to enhance transparency and accountability, concerns have been raised about the potential misuse of AI-generated content, especially with regards to influencing elections.
To address these concerns, twenty internet businesses, including TikTok, signed a contract earlier this year to combat the threat of AI manipulation in elections. As part of their commitment, TikTok will now use Content Credentials to label and identify content that was produced outside the platform’s domain. The company is currently in the process of assigning labels to material generated by AI within the app itself.
Adam Presser, head of operations, trust, and safety at TikTok, emphasized their policies against unlabeled realistic AI content. He stated, “If realistic AI (generated contents) appears on the platform, then we will remove it as a violation of our community guidelines.” This assertion highlights TikTok’s proactive approach in maintaining a safe and reliable user experience.
The development of Content Credentials technology can be credited to the Coalition for Content Provenance and Authenticity, formed by industry giants such as Adobe and Microsoft. OpenAI, the organization responsible for creating the ChatGPT tool, has already committed to using this technology. In fact, Meta Platforms (owner of Instagram and Facebook) and Google (owner of YouTube) have also expressed their willingness to implement Content Credentials. This collaborative effort among major tech companies demonstrates a collective commitment to combatting the spread of AI-generated misinformation.
For Content Credentials to effectively operate, there must be a standardized agreement between content distribution platforms and generative AI tool designers. OpenAI’s Dall-E tool, for example, adds data to generated images that can be used to assess whether they have been modified. These images are also embellished with a watermark by OpenAI. This way, when such images are uploaded to TikTok, they can be automatically identified as AI-generated content.
It is worth noting that TikTok is owned by Chinese company ByteDance, which currently faces regulatory challenges in the United States. To avoid potential market restrictions, ByteDance is required to sell TikTok. In response, both TikTok and ByteDance have filed a lawsuit against the government to prevent the perceived violation of the First Amendment.
In conclusion, TikTok’s introduction of Content Credentials is a significant step towards combatting the potential misuse of AI-generated content. By working alongside other tech giants and implementing this technology, TikTok aims to enhance transparency and maintain a safe environment for its users. While concerns about AI manipulation persist, these collaborative efforts reflect the industry’s commitment to addressing this issue and ensuring the integrity of online content.