In a significant move towards transparency and authenticity, OpenAI has joined forces with the Coalition for Content Provenance and Authenticity (C2PA) to label AI-generated videos. This collaboration aims to provide viewers with clear context, ensuring they recognize when content is the product of artificial intelligence. Let’s dive into the details of this groundbreaking development.

TikTok Takes the Lead

TikTok, the popular video-sharing platform, has taken the lead by automatically flagging AI-generated content (AIGC) uploaded from other platforms. Until now, creators could sometimes evade labeling rules by uploading AI-generated material from external sources. However, TikTok will now use digital watermarks provided by C2PA to identify and label AIGC effectively.

How Does It Work?

1. Labelling AI-Generated Content: TikTok’s own AI tools already label content created within the platform. Now, this labeling extends to AIGC from other sources. The goal is to prevent confusion or misinformation for viewers who might not realize that certain content was AI-generated.
2. Content Credentials Watermark: TikTok will use the Content Credentials watermark, a technology developed by C2PA, to identify AIGC. This watermark provides essential metadata about when, where, and how the content was created or edited. It ensures transparency and helps viewers understand the context
3. Two-Way Labelling: TikTok isn’t just labeling content; it’s also applying the same watermarking technology to videos downloaded from its platform. This means that other platforms can recognize TikTok-originated content as well. However, this labelling capability is limited to platforms that are also part of the C2PA coalition, including major players like Microsoft, Google, and Adobe.

OpenAI’s Role

OpenAI, known for its cutting-edge AI models, has stepped up to the plate. Although OpenAI wasn’t initially part of the C2PA, it recently joined the steering committee. OpenAI has already integrated the Content Credentials technology into its video-creation AI model, Sora. Once Sora is released to the public, it will automatically label the video clips it generates as products of AI.

While major players are embracing labeling and transparency, smaller AI groups may continue producing unmarked content. Open-source tools, like Stable Diffusion, can still create unlabelled AI content. However, initiatives like C2PA are crucial steps toward building trust and ensuring viewers understand the origins of the content they consume.

In summary, OpenAI’s commitment to labeling AI videos aligns with the industry’s push for transparency. As Sora prepares to make its debut, we can expect a new era of AI-generated videos with clear context and authenticity.