TikTok has announced more ways for people to spot, shape, and understand their experiences with AI-generated content (AIGC). Key updates include:
- Testing a new AI-generated content control in ‘Manage topics’ feature, to help people choose how much AIGC they want to see in their For You feeds.
- Bolstering AI-generated content labelling strategies with a technology called “invisible watermarking.” Over 1.3 billion videos have been labeled as AI-generated on TikTok to-date.
- Launching a $2M AI literacy fund and joining the Partnership on AI, in order to deepen TikTok’s industry collaboration and AI education efforts.
TikTok AI-generated content control
TikTok are to start testing a new AI-generated content control in their ‘Manage topics’ feature to empower people to choose how much AIGC they want to see in their For You feeds.
Manage Topics already enables people to adjust how often they see content related to over 10 categories, the AIGC setting is intended to help people tailor the diverse range of content in their feed, rather than removing or replacing content in feeds entirely. This means that people who love AI-generated history content can see more of this content, while those who’d rather see less can choose to dial things down.
AIGC labeling
TikTok are also testing a solution called “invisible watermarking.”
TikTok already require people to label realistic AI-generated content on TikTok and layer multiple strategies to apply that rule. This includes labeling tools that TikTok offer creators as well as our their detection models. TikTok also use a cross-industry technology called C2PA Content Credentials, which embeds metadata into content that lets TikTok — as well as other platforms who use C2PA — know when something is AI-generated. These efforts helped label over 1.3 billion videos to date.
However, a common industry challenge is that these kinds of labels may be removed when content is reuploaded or edited elsewhere. “Invisible watermarks” add another layer of safeguards with a robust technological “watermark” that only TikTok can read, making it harder for others to remove.
Deepening industry collaboration
TikTok are continue deepening industry partnerships to better advance AI transparency and education together. Now, they are launching a $2M AI literacy fund for experts like GirlsWhoCode to create For You feed content that teaches people about AI literacy and safety. It includes over 20 experts across over a dozen markets globally.
We continue to refine our labeling approach as industry norms and expert guidance evolve, while maintaining firm policies against harmful AIGC. For example, over the last year, we enhanced our AIGC labels by adding more context around whether content was labeled due to our AI detection, creator labels, or TikTok AI tools. We’ll continue to iterate as our industry evolves in the months to come.
– TikTok
