A landmark shift in how artificial intelligence is regulated arrived in Europe this week, as key transparency provisions of the European Union's AI Act came into full legal force.

The rules, now enforceable across all 27 EU member states, require companies to inform citizens when they are interacting with an AI system — whether that's a chatbot, an automated customer service voice, or an AI-powered recommendation engine. Critically, AI-manipulated or AI-generated content, including text, audio, images, and video, must now be clearly labeled as such.

Companies that fail to comply face significant penalties: fines of up to €15 million (approximately £12.8 million) or 3% of annual global turnover, whichever is higher. For global technology companies, that second figure could run into the billions.

Henna Virkkunen, an Executive Vice President of the European Commission, called the new enforcement milestone "an important step towards AI that people and businesses can understand and trust."

The transparency regulations are part of the EU's AI Act — the first comprehensive legislation of its kind anywhere in the world, passed after years of public debate as AI systems became rapidly more capable and more widespread. The Act is designed to place meaningful safeguards around the technology and to empower citizens to make informed decisions about when and how they engage with AI.

For the creative community, the rules represent a meaningful victory. Artists, writers, musicians, and filmmakers have spent years sounding the alarm about the flood of AI-generated content — sometimes called "AI slop" — that has been competing with human-made work on platforms and in search results. The new labeling requirements should help readers and viewers identify which content was made by a person.

"This is an important step," said one European filmmaker. "At least now people will know what they're looking at."

Not everyone is satisfied. Many creators had hoped the Act would go further and ban the practice of training AI systems on copyrighted work without compensation or consent — a concern these transparency rules do not address. That fight continues in European courts and at the legislative level.

There are also practical concerns about implementation. Boniface de Champris, from the Computer and Communications Industry Association, raised a pointed question: if mandatory labeling becomes so broad that everything from spell-checked emails to filtered photos must be labeled, will the labels mean anything at all? "If we have to label everything," he warned, "the labelling of AI content will lose all meaning."

These tensions point to the genuine complexity of regulating AI in a rapidly evolving landscape — one where the lines between human and machine-generated content blur a little more every month. But for now, the enforcement of these transparency provisions marks a concrete and historic milestone.

For the first time, people across Europe have a legal right to know when they are talking to an AI. And companies have a legal obligation — backed by real financial consequences — to tell them.

The broader AI Act, with additional provisions governing higher-risk AI applications in healthcare, employment, and law enforcement, will continue to roll out through 2027 as part of the law's phased implementation schedule. Today's enforcement milestone is the beginning of a much larger transformation in how AI and humans co-exist — one built, for the first time, on transparency.