Politics
TikTok rolls out deepfake detection tool for public figures to fight back against disinformation
Key Points
The social media company’s ‘Likeness detection’ feature is aimed at politicians dealing with the deluge of disinformation online. TikTok announced on Friday that it had begun rolling out an opt-in feature called 'Likeness Detection' to help creators find and flag AI-generated content that uses their likeness on the platform, as the company tries to fight back against AI-generated disinformation. The tool is aimed at helping public figures, including members of government, politicians and...
The social media company’s ‘Likeness detection’ feature is aimed at politicians dealing with the deluge of disinformation online.
TikTok announced on Friday that it had begun rolling out an opt-in feature called 'Likeness Detection' to help creators find and flag AI-generated content that uses their likeness on the platform, as the company tries to fight back against AI-generated disinformation.
The tool is aimed at helping public figures, including members of government, politicians and political party accounts, to spot and report AI-generated content that uses their image without consent.
It comes at a time when Europe is already facing a wave of AI-driven disinformation and deepfake campaigns disrupting its election processes, and with nine more European countries heading to the polls in the next 12 months, it could prove decisive.
Eligible users can sign up by verifying their ID with a trusted third-party partner. They are then notified when content and accounts on TikTok may be using their likeness and can review them within the tool.
Users can then report content directly to TikTok’s safety teams for review and action.
The EU began rolling out artificial intelligence transparency rules on 2 August under the AI Act, including a requirement for deepfakes and other AI-generated content to be labelled.
TikTok has made efforts to curb deepfakes in recent years, such as requiring content creators to label realistic AI-generated images, audio and video as stipulated in the AI Act as well.
Identify AI-Generated Content
The company also launched a new transparency tool for users to check whether content is made or edited with the social media platform’s creation features.
The ‘Identify AI-Generated Content’ tool is free to use by anyone, said TikTok, even those without an account.
The platform allows users to upload an image or video to check if it was generated or edited with TikTok’s own AI tools including parent company ByteDance’s CapCut, Dreamina, Hypic and Pippit apps.
It supports a wide range of formats including popular ones like JPEG, PNG, and MP4, with a file size limit of 20 megabytes per image or 100 megabytes for each video or audio.
The tool checks for two signs that the content was AI-generated: metadata credentials added by TikTok and ByteDance apps and invisible watermarks embedded directly into the content that are harder to remove.
The company has said that the tool deletes all content as soon as it has been uploaded and checked, and it is never used to train AI models.
TikTok warn users that it only detects whether AI edited or created content using its own or ByteDance app tools, so it’s not a foolproof way to spot AI content.