AI Act label
Add the EU AI label to your images and videos
Since 2 August 2026, the EU AI Act asks whoever publishes a deepfake to say that AI made or changed it. Drop your files, place the official EU icon and download them labelled.
Your files
1 credit per file. 2 free files a day without an account. Files deleted after 24 hours.
How it works
Drop your files
Images and videos, up to 50 at once. You see the cost before anything starts.
Place the icon
Pick one of the three EU icons, then move it, resize it and check it on each file.
Download them labelled
Each file comes back with the icon burned in. Images take seconds, long videos a few minutes.
What the AI Act asks
Article 50 of the EU AI Act sets transparency rules for AI. Its fourth paragraph is about what you publish: if you share an image, a video or a sound that AI made or changed, and it could pass for real, you have to say so.
The label must be clear and visible the first time people see the content, and accessible to everyone (Article 50(5)).
These rules apply since 2 August 2026 (Article 113). Breaking them can cost up to EUR 15 million or 3 % of worldwide yearly turnover, whichever is higher. For a small or medium business, whichever is lower (Article 99).
Deployers of an AI system that generates or manipulates image, audio or video content constituting a deep fake, shall disclose that the content has been artificially generated or manipulated.
Do you need to label your content?
In the law, a deepfake is an image, a sound or a video made or changed by AI that looks like real people, objects, places or events and could pass for real (Article 3(60)).
Label it
Realistic images and videos
A face, a place or an event that could pass for real. Label it where people see it first.
Label lightly
Art, satire and fiction
Still labelled, in a way that does not get in the way of the work.
Label, unless reviewed
Text that informs the public
AI text on matters of public interest is labelled, unless a person reviewed it and answers for it. This page labels images and videos only.
No label
Uses allowed by law against crime
Uses authorised by law to detect, prevent, investigate or prosecute crime need no label.
Which EU icon to choose
The European Commission made three icons free for everyone to use, without naming the source. Using them is optional; saying that AI was involved is not.
Where to put the icon
The Commission's guidance, in four points.
Seen at first sight
People should see the label as soon as they see the content, not at the end.
Clear of the app's buttons
The label must not hide under buttons or captions. On a vertical video for TikTok, Reels or Shorts, the right side and the bottom are covered: put it at the top.
In the picture itself
Burned into the pixels, the label stays when the file is shared, downloaded or posted again.
Big enough to read
Pick a clear size. When you post, add alt text that says the content is AI-generated.
UncovAI detectors
Not sure it is AI? Check it first.
Our detectors tell you in a few seconds if an image or a video was made by AI.
Sources: Regulation (EU) 2024/1689, the AI Act (Articles 3, 50, 99 and 113), and the European Commission's page on the EU icons for labelling AI-generated content.
Frequently Asked Questions
How does UncovAI detect AI-generated images?
UncovAI analyzes visual artifacts, texture inconsistencies, and generation fingerprints often left by AI models.
Can it detect images from Midjourney, DALL·E, Flux, Grok, or similar tools?
Yes. The detector is built to identify patterns from major image generation models and is updated regularly.
Do you store my image?
UncovAI does not store, log, or share any image, video, audio, or text submitted for detection. All content is processed in real time and discarded immediately after analysis. This zero-storage policy makes UncovAI safe for sensitive use cases including legal evidence, private communications, corporate security screening, and personal privacy protection.
Can real photos be flagged as AI?
Sometimes. Smartphone post-processing, strong filters, or heavy editing can look synthetic to detection systems.
What image formats are supported?
Most standard formats are supported, such as JPG, JPEG, PNG, WEBP, and HEIC (within size limits).
Can this help with scam, deepfake or disinformation checks?
Yes. It’s useful as a first authenticity signal, but sensitive decisions should include manual review and extra verification.
How accurate is UncovAI's AI image and deepfake detector?
UncovAI provides high-accuracy AI detection across major image generators and deepfake tools. However, no detector — including UncovAI, Google’s SynthID, Decopy AI, TruthScan, or WasItAI – achieves 100% accuracy. Results are presented as probabilistic confidence scores rather than binary verdicts. Accuracy is affected by:
- Post-processing (filters, compression, resizing)
- Smartphone AI photo enhancement (computational photography)
- Heavily edited hybrid images
UncovAI recommends treating all results as an advisory signal and applying human review for high-stakes decisions.
Feel free to get in touch
with UncovAI
If you’re intrigued by our technology or if you’re interested
in exploring potential partnerships!