EU's new AI transparency rules take effect, granting new powers to Commission

The European Union has implemented new transparency rules for artificial intelligence systems, compelling them to inform users when they are interacting with a machine or when content has been altered. These regulations, effective Sunday, also empower the European Commission's AI Office with enhanced authority to oversee generative AI models and levy fines for violations.

Companies like OpenAI, Google, and Anthropic must now ensure their systems clearly indicate machine interaction. This applies to AI used in public services, customer interactions, and fraud reporting. Systems analyzing emotions or using biometric categorization must also provide warnings, with exceptions for law enforcement purposes.

Starting December 2, AI systems will be required to watermark AI-generated text, photos, videos, and audio to help users identify machine-created content. However, platforms like ChatGPT and Gemini will not be legally liable for user-generated content; users themselves must disclose manipulation. Fictional, artistic, or satirical content is exempt, but AI-generated information on public interest topics like elections or public health must be disclosed unless human-supervised.

Further prohibitions include the ban on AI-generated sexualized images without consent, a measure prompted by controversies surrounding AI-generated images. Platforms like X's Grok tool have until year-end to implement protective measures. The AI Office's new powers are designed to ensure compliance and address emerging challenges in AI deployment.