European Union regulations stemming from the AI Act will require companies to label artificially generated images, audio, and text that are designed to appear authentic, with compliance beginning 2 August for new AI systems. Existing systems have an additional four months to adhere to the requirements. The rules mandate that synthetic media be visibly marked as AI-generated and include digital watermarks indicating artificial origins. Content on matters of public interest must also be labeled as AI-produced when lacking human editorial supervision.
Under the new framework, companies have the option to use black-and-white labels created by the EU executive or design their own labeling systems. Companies are encouraged, though not required, to retroactively label AI-generated content created prior to the rules’ implementation. Exemptions exist for personal content and works clearly identified as artistic, satirical, or fictional. Non-compliance could result in financial penalties reaching €15 million or 3% of a company’s annual worldwide revenue.
The regulations were developed in response to the proliferation of deepfakes and manipulated media depicting politicians and public figures engaged in various activities, including alleged election-rigging conspiracies and fabricated speeches. The initiative reflects concerns about protecting both consumers and democratic processes by ensuring transparency regarding AI-generated content.
Industry representatives have expressed support for combating deepfakes but raised concerns that the rules may be interpreted too broadly, potentially resulting in excessive labeling that could desensitize users. Technology associations argue that recent EU guidance has expanded the definition of deepfakes beyond the original 2024 AI Act framework, potentially affecting advertising, publishing, and film sectors where AI use is already widespread.
Several major technology companies have already implemented or announced AI transparency measures. Meta and Google have signed a voluntary code of practice intended to facilitate compliance, with more than 180 organizations joining the initiative. TikTok reports having labeled over 3 billion pieces of content, while Google’s SynthID tool has watermarked more than 100 billion images and 60,000 years of audio content.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI