The European Union has officially started the clock on one of the most consequential AI disclosure rules in history. From this week, the EU AI Act requires anyone deploying AI systems that generate or manipulate deepfake content to clearly label it as synthetic.
The formal obligation under Article 50 of the AI Act takes full effect on August 2, 2026, though transitional relief for older generative AI systems extends until December 2, 2026. Non-compliance carries serious financial consequences, with fines reaching up to €15 million or 3% of total worldwide annual turnover, whichever is higher.
What the rule actually requires
Article 50(4) is direct: deployers of AI systems that produce deepfakes must disclose the synthetic nature of that content to avoid misleading the public. The European Commission published draft guidelines supporting these transparency obligations in May 2026, with final guidelines expected around July 20, 2026. A Code of Practice specifically covering marking and labeling of AI-generated content was announced in June 2026, giving companies a framework to work toward before the hard deadline arrives.
New rules targeting non-consensual intimate deepfakes, agreed upon in 2025, are set to begin enforcement in late 2026, adding another layer of accountability to the synthetic media ecosystem.
