TikTok Launches AI Transparency and Portrait Detection Tools
TikTok announced a transparency tool that lets anyone, including users without an account, upload an image or video to check whether it was generated or edited by TikTok's AI features. The tool checks content credentials and the proprietary VeraDura invisible watermark, while the opt-in Portrait Detection feature helps eligible creators find and flag AI content that may use their likeness without authorization. TikTok also says it has labeled more than five billion AI-generated videos and removed over 99% of violating AI content before user reports, figures that IT之家 AI attributed to the company without independent measurement.
The launch adds public AI-content verification and opt-in likeness detection to TikTok's platform-level safeguards.
IT之家 11 月 22 日消息,当地时间 11 月 19 日,TikTok 宣布推出两项新工具,进一步推进人工智能透明度建设,让社区用户在 TikTok 上创作和接触人工智能内容时,拥有更高的知情度与掌控力。
TikTok 推出了一款新的透明度工具,让任何人都能自主进行核验。即便没有 TikTok 账号,用户也可以上传图片或视频,检测其是否由 TikTok 的人工智能创作功能生成或编辑。除了内容凭证外,该工具还会检测 TikTok 专有的隐形水印 VeraDura—— 该水印持久性更强,也更难以被去除。

此外,TikTok 将面向全球逐步推出“肖像检测”功能。这项需用户主动开启的工具,能够帮助符合条件的创作者在 TikTok 上发现并标记可能未经授权使用其肖像的人工智能生成内容。
IT之家从官方新闻稿获悉,迄今为止,TikTok 已为平台上超过 50 亿条 AI 生成视频添加标识;违规 AI 内容中超过 99% 在被用户举报前即被平台主动移除。
