PhotoGuard is MIT research (MadryLab, MIT licence) that protects photos from AI editing. Its code needs Python and a big GPU. Photo Shield runs its method in your browser.
What is the same and what is different
- Same idea: PhotoGuard-style L-infinity PGD (projected gradient descent) with an invisible pixel budget. Photo Shield attacks the encoder and decoder together, so the AI's copy of your photo comes out as broken as possible.
- Different encoder: it attacks TAESD (MIT licence), a small copy of the Stable Diffusion image encoder, so it fits in a browser. This makes it weaker than running PhotoGuard on the full model.
- No setup: no Python, no GPU server and no upload. It is fastest on browsers with WebGPU (recent Chrome and Edge).
Photo Shield is not made by or affiliated with MIT or MadryLab. It is an independent tool built on their open-source research.
How to use it (3 steps)
- Open Photo Shield and choose a JPG, PNG or WEBP photo. It stays on your device.
- Leave the bars on full protection (the default) and press Protect my photo. It takes about 1-2 minutes on a recent laptop.
- Check the protection score, then download the protected PNG and post that version.
Honest limits
- It is aimed at AI photo editors built on Stable Diffusion. It does not reliably stop ChatGPT, Gemini or face-swap apps.
- Screenshots, heavy compression and resizing weaken it. Post the downloaded PNG where you can.
- No tool can make a photo impossible to edit. Combine it with private accounts and posting fewer close-up face photos.
More guides
- Anti-AI photo filter
- Protect images from AI
- Make an AI-proof image
- Stop AI editing my photos
- Protect your face from AI
- How to stop AI using my photos
- PhotoGuard online
- Protect photos from deepfakes
- Protect Instagram photos from AI
🛡️ Protect a photo now - free, in your browser
Last updated: October 2026