Starting August 2, 2026, the European Union requires companies to label content made by artificial intelligence and to tell people when they are talking to an AI chatbot instead of a real person. The rules apply in all 27 EU countries.

The European Commission announced the start of enforcement on July 31, 2026. These transparency rules are part of the EU's larger AI Act, which also covers high-risk AI systems and general-purpose AI models — the large, versatile systems that power tools like chatbots and image generators.

The Commission published guidelines on how to handle AI-generated content on July 29, 2026, five days before enforcement began. It also maintains a Code of Practice to help companies follow the rules.

The rules use two levels of labeling. If a machine helped create something that looks real but is fake — like a video showing a politician saying something they never said, or a fake audio clip — it needs an "AI" mark. If content was fully created by AI, such as AI-composed music, AI-generated art, or AI-written news summaries, it needs an "AI-Generated" label. AI systems that read emotions or analyze biometric data, like facial recognition, also need labeling.

Some things are exempt. Personal content like group chats is not covered. Clearly artistic or satirical works are also exempt, which should help artists and creators avoid extra paperwork. The AI Act does not regulate AI systems that pose minimal or no risk.

For companies that build general-purpose AI models, the European Commission will directly oversee them. These companies must keep documentation, publish summaries of what data they used to train their models, and have copyright policies in place. Individual countries will enforce the labeling rules for everyone else.

The penalties are steep. Companies that break the transparency rules can be fined up to €15 million or three percent of their global yearly revenue, whichever is higher. EU institutions face a lower cap of €750,000.

Not everything in the AI Act is on schedule. EU governments and lawmakers agreed in May 2026 to delay some rules for high-risk AI systems, including those used in biometrics and critical infrastructure. A retail group argued in June 2026 that AI-generated ads should be exempt from the labeling rules. Neither of those efforts changed the Article 50 obligations that are now in effect.

For technology companies operating in the EU, the requirements are immediate. Any platform that shows users AI-generated text, images, audio, or video must add labels that both computers and people can read. Chatbots and AI agents must disclose that the user is talking to a machine, not a person. Companies building general-purpose AI models must maintain documentation, publish training data summaries, and have copyright policies ready.

The transparency rules sit alongside the AI Act's broader system for managing AI risk. High-risk system rules, general-purpose AI model rules, and now the transparency rules form three connected layers of regulation. Companies that already set up systems to follow the high-risk rules will find the transparency rules add another layer on top. Those that have not started now face all three at once.

One open question is how strictly the rules will be enforced. Each EU country is responsible for enforcing compliance, and some may be more active than others. The European Commission directly oversees the largest AI model providers, which keeps enforcement centralized for the biggest systems. For the content labeling rules, the real test will be whether national regulators focus on companies that ignore the rules broadly rather than going after individual violations.

The broader context here is that the EU is the first major government to put binding, enforceable AI transparency rules in place at this scale. Think of it like nutrition labels on food packaging — the EU is essentially saying that if content was made by a machine, people have a right to know. Any company that serves users in the EU will need to meet these standards, no matter where they are based. Whether other countries adopt similar rules, go their own way, or create conflicting ones will shape how AI-generated content is handled worldwide. Companies building AI tools that create or display content should expect that labeling will become a standard expectation everywhere, not just in Europe.