Under EU regulations, AI labels are now mandatory for content that appears authentic.

      Companies that create or utilize AI will soon be required to label content that appears authentic but is not, as the European Union's significant AI Act transitions from conceptualization to enforcement. Starting August 2, providers and users throughout the EU must adhere to binding transparency regulations that address deepfakes, synthetic media, and machine-generated text—a change that platforms have been preparing for since YouTube implemented auto-labeling for AI videos.

      The fundamental principle is straightforward. If content is capable of being mistaken for genuine, individuals have the right to be informed that it was created by a machine, and the responsibility now lies with the companies involved.

      The regulations divide the responsibilities into two categories. Providers, or the companies that develop generative systems, are tasked with embedding machine-readable identifiers to ensure that images, audio, video, and text can be recognized as AI-generated or altered. Deployers handle the visible aspect; anyone utilizing AI to generate a deepfake or to publish machine-generated text on public topics must transparently disclose that information to the audience in an clear and timely manner.

      Intent does not exempt anyone from responsibility. The obligation to label a realistic deepfake applies regardless of whether the creator intended to mislead, effectively closing a loophole that might have been exploited with claims of "we were only joking."

      Chatbots are also included in the regulations. Individuals engaging with an AI system must be informed at the onset of the interaction, rather than in the fine print of a terms-of-service document that often goes unread.

      There are specific exceptions. Content that is clearly unrealistic or fantastical is not subject to these rules, and AI-generated text does not need to be labeled if it has been meaningfully reviewed by a human with genuine editorial oversight.

      Creative works are given a lighter approach. Artistic and satirical content must still include a disclosure, but it should be presented in a manner that doesn't detract from the overall experience of the work, acknowledging the needs of filmmakers and artists.

      The penalties associated with these rules are substantial. Violations could result in fines of up to €15 million or 3% of a company's global annual revenue, whichever is greater, providing a strong incentive for compliance among businesses in Silicon Valley and Brussels.

      The implementation timeline is somewhat flexible. The obligations officially commence on August 2, but a separate simplification package has been proposed that could extend the deadline for machine-marking to December, as part of a broader effort to streamline certain aspects of the AI Act.

      The industry has been actively advocating for changes. Advertisers and others have argued that AI-generated advertisements should be exempt from the transparency requirements, a lobbying effort that has proceeded alongside the law's implementation.

      The Commission has sought to clarify the regulations. In May, it released draft guidelines and is finalizing a code of practice on how to label and mark AI content, which companies can endorse to demonstrate their compliance.

      The scope of these regulations extends beyond Europe. Because global platforms typically do not create separate systems for individual markets, the EU's labeling framework is likely to influence global practices for marking AI content, similar to the impact of its privacy legislation.

      Enforcement will pose significant challenges. Watermarks can be removed, metadata can be lost, and identifying machine-generated text is notoriously difficult; therefore, the efficacy of the rules will depend on the effectiveness of the tools designed to enforce them.

      Deepfakes represent the most serious challenge. The EU has taken steps to prohibit non-consensual intimate deepfakes, and the labeling framework aims to address the broader range of synthetic media to prevent the complete erosion of trust.

      The implications extend beyond mere compliance. Europe is wagering that mandatory transparency can help maintain a collective understanding of what is real online, at a time when generative tools make the creation of convincing fakes exceedingly easy.

      Whether this approach will succeed remains uncertain. The legislation represents the most ambitious effort to label synthetic content to date, but its effectiveness hinges on the ability to detect such content keeping pace with its generation, a contest that the creators of fakes have so far been winning.