欧盟根据《AI法案》规定，自2025年8月2日起将强制要求公司对具有真实感的AI生成内容贴上数字水印标签[1]。这一规则涵盖AI生成的图像、音频和文本内容[1]。新上市的AI系统需要立即合规，而现有系统可获得为期4个月的宽限期[1]。违反该要求的企业将面临最高达其总营收3%的罚款[1]。
该规则设有豁免条款，个人内容、艺术性讽刺和虚构作品不适用标注要求[1]。欧盟议会议员谢尔盖·拉戈金斯基表示："这不仅是消费者保护问题，也是民主保护问题。"[1]谷歌旗下的SynthID已标记超过1000亿张图像和6万年时长的音频内容[1]。
The European Union is implementing a requirement for companies to label artificial intelligence-generated content that resembles authentic material, effective August 2, 2025, under the provisions of the AI Act [1]. The mandate applies to AI-generated images, audio, and text that display realistic characteristics, with companies required to apply digital watermark labels to such content [1].
New AI systems must comply with the labeling requirement immediately upon deployment, while existing systems have been granted a four-month grace period for implementation [1]. Violations of the rule will result in fines reaching up to 3 percent of a company's total annual revenue [1]. Certain categories are exempt from the labeling requirement, including personal content, artistic satire, and fictional works [1].
EU Parliament member Sergey Lagodinsky emphasized the regulatory purpose, stating: "It is a matter not only of customer protection, it's also a matter of democracy protection" [1]. Google's SynthID tool has already labeled over 100 billion images and 60,000 years' worth of audio content, demonstrating existing technological capacity to implement such measures [1].