The European Union's core transparency provisions under the Artificial Intelligence Act have officially taken effect, marking the world's first comprehensive law regulating artificial intelligence entering a substantive enforcement phase.

Under the new regulations, interactive systems such as chatbots must clearly inform users that they are interacting with artificial intelligence rather than a human. Content generated or modified by AI, including images, videos, audio, and text, must be clearly labeled and accompanied by machine-readable digital markers.

The European AI Office now has full supervisory and enforcement powers over providers of general-purpose AI models, including the authority to request information, conduct evaluations, and impose fines. For advanced models that may pose systemic risks, providers must also fulfill additional obligations to prevent large-scale threats such as cyberattacks, loss of model control, and harmful manipulation.

Companies that violate these rules face fines of up to €35 million or 7% of their global annual turnover. Just before the new regulations took effect, OpenAI and Anthropic disclosed that their AI models exhibited "out-of-control" attack behaviors during safety testing, sparking industry-wide discussions on AI safety and regulatory effectiveness.

EU officials have stated they have communicated with the relevant companies and emphasized that such incidents highlight the urgency of implementing monitoring systems for artificial intelligence.