# Do you have to label AI videos? The EU deepfake rule explained > The EU AI Act asks people who share realistic deepfakes to say they are AI-made. Here is what that means for a fun face swap video and how to label it. Source: https://vswp.app/guides/label-ai-video-eu-ai-act/ · Updated 2026-10-29 In the European Union, in most cases yes. The AI Act's transparency rule, which applies from August 2026, says that whoever uses AI to make a deepfake (a realistic video, image or audio of people or events that never happened) has to disclose that it was artificially generated or changed. For clearly creative, funny or fictional work the duty is lighter but doesn't disappear, so the simple habit that works almost everywhere is to write "made with AI" in the caption and turn on the platform's AI label. This is not legal advice. ## What does the EU rule actually say? The AI Act is the European Union's law on artificial intelligence. Most of it is about the companies that build and sell AI systems, but one part, Article 50, is about being open with the public, and its rules apply from August 2026. For deepfakes it sets two kinds of duties: - **For the makers of AI tools:** what their tools generate should be marked in a machine-readable way, so software can tell it's AI-made. - **For the people who use the tools to share a deepfake:** they have to disclose that the content was artificially generated or manipulated. The disclosure has to be clear and reach viewers no later than the moment they first see the video. A note buried somewhere in your profile doesn't do that job. ## What counts as a deepfake? The Act describes a deepfake as AI-made or AI-changed video, image or audio that looks like real people, places or events and could falsely seem authentic to someone. A realistic AI person swap, where your face shows up in a music video you were never in, fits that description well. A cartoon filter nobody could mistake for real footage usually doesn't. Our [AI video glossary](/guides/ai-video-glossary/) explains the related terms. ## Does it apply to a fun video with friends? Two details matter here. **Creative and satirical work.** When a deepfake is part of an evidently creative, satirical, artistic or fictional work, the duty is limited: you still say AI is involved, but in a way that doesn't spoil the work. A parody still gets a label; the label just doesn't have to kill the joke. **Purely personal use.** The Act's duties for users are aimed at professional use, and purely personal, non-professional activity is treated differently. A clip kept in a family chat is not the same situation as a public post. But once a video is public, the platforms' own rules apply anyway, and the line isn't always obvious: an account with sponsors or a business page is not personal. So the practical answer stays the same: label it. ## Who labels: you, the app or the platform? Often all three play a part, and none of them replaces the others. - **The app** may add a visible mark or hidden information to the file. - **The platform** may ask you to switch on an AI label, or add one itself when it detects AI content. - **You** decide how the video is presented: the caption, the context and who sees it. Hidden information in a file can get lost when a video is edited, compressed or re-uploaded. Your own words in the caption are the label that reliably stays with the post. ## How do you label an AI video? 1. **Say it in the caption.** Put "made with AI" or "AI face swap" near the start, where people read it before they scroll on. 2. **Turn on the platform's AI label.** Our guide to [posting an AI video on TikTok, Reels and Shorts](/guides/post-ai-video-tiktok-reels-shorts/) shows where to find it. 3. **Label videos without a caption too.** In a messenger, say it in the message, or add on-screen text such as "AI video". 4. **Leave marks alone.** Don't crop, blur or cover marks when you repost, yours or anyone else's. 5. **Match the label to the risk.** If a viewer could believe the scene really happened, make the label more prominent, or don't post it. ## What about outside the EU? Rules differ by country. Some countries have rules on AI content or on using a person's likeness, others rely on general laws about fraud, defamation or privacy, and many are still writing new ones. Big platforms also apply their AI labeling rules worldwide. If you're not sure what applies to you, label anyway: it costs nothing and protects the people in the video. ## Does a label replace consent? No. A label tells viewers what they're watching; it doesn't give you permission to use someone's face. Ask before you use anyone's photo, and ask again before a private joke becomes a public post. Our guide to [face swap consent and safety](/guides/face-swap-consent-and-safety/) covers how to ask and what never to make. ## How does this work on vswp? On [vswp](/) you pick a video from Trends or upload your own, add a photo for each person to replace, and get a new video where the moves, camera, lighting and sound stay as in the original. The free preview carries a small "vswp.app" mark. Full videos, paid in Telegram Stars (⭐), have no mark, so the label is up to you when you share one. By pressing "Create video" you confirm you have the right to use the video and the photos. And if you want to see your clip the way a skeptical viewer would, the signs are in [how to spot a deepfake](/guides/how-to-spot-a-deepfake/). ## FAQ ### Do I have to label an AI face swap video in the EU? If it's a realistic deepfake you share, the AI Act's disclosure rule, applying from August 2026, generally says yes. Rules differ elsewhere, and this is not legal advice. ### Is a parody exempt from the label? Not fully. For evidently creative, satirical or fictional work the duty is lighter, but you still say AI is involved, in a way that doesn't spoil the joke. ### Is the platform's AI label enough? It's a good start, but it may not show everywhere the video travels. Add "made with AI" to the caption too, so the label stays with the post. ### Do I need to label a video I only send to friends? Purely personal use is treated differently by the Act, but friends forward videos. Saying "AI video" in the message is easy and keeps a private joke from turning into a misunderstanding.