欧盟根据《AI法案》推出新规则，从2024年8月2日起，在欧盟市场上的新AI系统必须为人工生成的图像、音频和文本加上标签标记，以表明其人工来源。[1]现有系统获得额外四个月的合规期限。[1]违反这一要求的公司将面临最高1500万欧元或其全球营业收入3%的罚款。[1]
业界已开始采取行动。谷歌表示已在超过1000亿张图像和6万年音频上加上隐形数字水印，[1]而TikTok称已帮助标记超过30亿条内容。[1]此外，超过180个组织已签署由欧盟AI办公室监督的自愿性实践准则。[1]
欧盟AI委员会主任Henna Virkkunen指出AI是"具有变革性的技术"，但最先进模型"造成全新规模的风险"。[1]绿党议员Sergey Lagodinsky则强调，这项举措"不仅是消费者保护问题，也是民主保护问题"。[1]
The European Union has implemented mandatory labeling requirements for artificially generated content under its AI Act framework, effective August 2, 2024 [1]. All new AI systems operating in the EU market must now clearly mark artificially generated images, audio, and text to indicate their artificial origins [1]. Existing systems have been granted an additional four-month compliance period to adapt to these requirements [1].
Companies failing to comply face substantial penalties, with fines reaching up to 15 million euros or 3 percent of global annual turnover, whichever is higher [1]. Henna Virkkunen, head of the EU AI Office, characterized AI as a "transformative technology" while warning that the most advanced models "create risks on an entirely new scale" [1].
Industry players have begun taking steps toward compliance. Google has already applied invisible digital watermarks to over 100 billion images and 60,000 years of audio content [1]. TikTok reports having assisted in labeling more than 3 billion pieces of content [1]. Beyond individual company efforts, over 180 organizations have signed voluntary industry practice guidelines overseen by the EU AI Office [1]. Green Party parliamentarian Sergey Lagodinsky framed the initiative as extending beyond consumer protection, stating: "This is not only a consumer protection issue, but also a democracy protection issue" [1].