EU AI Act Rules on AI Models Enter Force, Targeting Transparency and Deepfakes
Brussels moves from theoretical oversight to active enforcement as core disclosure duties for AI systems become legally binding.
The European Union has officially transitioned from drafting AI policy to active enforcement, with specific rules of the EU AI Act regarding AI models becoming enforceable on August 2, 2026. This milestone marks a critical shift in the global regulatory landscape, establishing Brussels as a primary authority for artificial intelligence safety and ethics.
Under the new enforcement phase, core disclosure duties are now legally binding. These mandates require that users be explicitly informed when they are interacting with an AI system. Furthermore, the rules require the clear labeling of deepfakes to prevent deception. While much of the broader regulation targets 'Big Tech' and the providers of general-purpose AI, these transparency requirements extend to smaller entities and everyday users who deploy AI for public-facing content.
A Risk-Based Framework
The EU AI Act represents the world's first comprehensive legal framework for artificial intelligence. It is built upon a risk-based approach, meaning the intensity of the regulation scales according to the level of risk a system poses to society. Rather than a single rollout, the EU has implemented a phased application of the law, allowing different provisions to become applicable at different intervals to give the industry time to adapt.
Implications for the AI Ecosystem
This shift to active oversight is designed to curb the spread of misinformation and ensure transparency across the entire AI ecosystem. By holding not just the model creators but also the deployers accountable, the EU is attempting to create a trustworthy environment for AI adoption. The move signals that the era of unregulated AI deployment in Europe has ended, forcing companies to integrate transparency into their product design by default.
Remaining Deadlines
While core disclosures are now in effect, some technical requirements follow different timelines. Under Regulation (EU) 2026/1744, known as the 'Digital Omnibus on AI,' the deadline for machine-readable watermarking obligations for AI-generated content—specifically for systems already present on the market—has been set for December 2, 2026. Industry observers will now be watching how strictly the EU enforces the August 2 disclosure mandates and how providers manage the transition to the December watermarking deadline.