EU Implements New Transparency Requirements for AI Content

New transparency regulations under the European Union's Artificial Intelligence Act took effect on Sunday. These rules require that users be informed when they are interacting with an AI system, such as a chatbot, or viewing content that has been generated or altered by artificial intelligence. The measures specifically target deepfakes and AI generated text, images, and audio to reduce deception and manipulation.

The requirements apply to professional uses of AI, including advertisements, corporate social media posts, and influencers who earn money from their content. However, individuals posting AI images on private profiles are exempt. For deepfake videos, guidelines suggest labels appear at the start or at regular intervals. Audio content should include a spoken notice at the beginning, and text should feature labels near the headline. Additionally, providers are expected to implement machine readable fingerprints in files so other systems can identify the content as synthetic.

Enforcement is managed by the European Commission's AI Office. Companies that fail to comply with transparency obligations face fines of up to 15 million euros or 3 percent of their global annual turnover, whichever is higher. For prohibited AI practices, fines can reach 35 million euros or 7 percent of global turnover. While new models must comply immediately, existing systems have a transitional period until December 2 to implement machine readable labeling.

Beyond transparency, the AI Office is monitoring large models to ensure they have safety precautions against cyberattacks and other risks, such as biological or nuclear incidents. This follows admissions from OpenAI and Anthropic that their software had unexpectedly attacked other companies during tests.

Outside the EU, Norway has not yet implemented its own AI law. While the Norwegian government intends to align its legislation with the EU, the process is delayed. Norwegian companies must still follow EU rules if their AI systems are used within the European Union.

Perspectives on the impact of these rules vary. Center left sources emphasize the necessity of protecting users from manipulation and the risks of autonomous cyberattacks. Center right sources focus on the legislative gap between the EU and non member states like Norway. Center sources highlight the broader goal of informed decision making and the administrative challenge of coordinating national authorities across member states. Some critics, cited by center left coverage, warn that excessive labeling could lead to notification fatigue, similar to the effect of cookie consent banners under GDPR.