Explainer: EU gets new powers over powerful AI from Sunday

Brussels, August 1, 2026 - From this Sunday, the European Union will gain the authority to enforce comprehensive artificial intelligence regulations, and companies that violate these laws will face substantial fines. The EU possesses one of the most extensive digital rulebooks globally, encompassing social media, e-commerce platforms, and search engines, with additional regulations slated for implementation throughout the year.

The AI Act, adopted in 2024, has been implemented gradually. Beginning this Sunday, the EU's risk-based approach to artificial intelligence will come into force. Companies will face increasing obligations depending on the degree of risk to European individuals' rights or health. This includes stricter requirements for firms developing AI technologies used for predictive policing, emotion recognition systems in workplaces or schools, and AI designed to manipulate human behavior.

Starting this Sunday, AI-generated content, such as deepfakes, must be labeled, while companies must ensure that their AI systems, such as chatbots, disclose to users that they are interacting with AI. Existing AI systems will have until December 2 to comply with the new regulations, with exceptions for artistic, creative, satirical, or fictional work.

European regulators, housed within the AI Office, will assume enforcement powers from Sunday. This regulatory body, comprising numerous technology experts, lawyers, and economists, will be responsible for testing advanced general-purpose AI models, such as chatbots, to ensure compliance with the law. Companies must also provide the EU with access to their models, and in some cases, the EU may demand access even before the deployment of the model on the market.

The EU's newfound powers aim to address issues previously encountered, such as the struggle to access American AI developer Anthropic's Mythos model this year. With the rules set to take effect on Sunday, such challenges should no longer arise. National authorities will also be empowered to enforce the regulations for smaller AI systems.

In the event that the EU identifies companies breaching the rules, it has the authority to compel them to remedy the breach and impose fines if they fail to comply. The most severe penalties apply to businesses enabling their systems to engage in prohibited practices, with fines reaching up to seven percent of a company's annual global turnover or 35 million euros ($40 million), whichever is higher.

Other violations may result in fines of up to three percent of a company's annual global revenue or 15 million euros. Companies could also be penalized for obstructing or failing to cooperate with EU investigations.

Future regulations under the AI Act will become enforceable later this year and beyond. Beginning in December, there will be a ban on AI systems generating sexually explicit deepfakes, a response to widespread outrage over non-consensual explicit images generated by Elon Musk's chatbot, Grok. Additionally, the EU is postponing the implementation of some high-risk AI rules concerning potentially dangerous models to safety, health, or citizens' fundamental rights, providing companies more time to prepare.

These rules will apply to standalone high-risk AI systems starting December 2027 and embedded AI tools in products from August 2028.