Noisify adds invisible adversarial noise to personal photos, making them resistant to AI-powered non-consensual image manipulation.
Noisify adds invisible adversarial noise to personal photos, making them resistant to AI-powered non-consensual image manipulation.
Project Details
Updated 07/08/26 · Provided via application · VerifiedNoisify is a web service for protecting personal photos from AI undressing without consent.
The path: a user uploads a personal photo to the service, chooses a protection level (this determines how visible the noise will be and how resistant the protection will be to bypassing and undressing). Then adversarial noise is applied to the photo. The user downloads the photo with the noise. If an attacker takes this photo and tries to undress it through AI tools, the result is destroyed and becomes unusable.
We use open-source protection tools and build an ensemble with our own modifications: Mist, z-image, CLIP ViT-L/14, with SD, Pixart, Flux planned, to maximize robustness across different models.
What exists now: first inference code is written, goal is the first protected photo.
What is planned: a working beta with verified protection efficacy against at least three major undressing tools.
Theory of Impact
Updated 07/08/26 · By grantmaking.aiAI undressing is a new form of gender-based violence, but with a much larger scale and possibilities for any person. There are too many such services, they allow either undressing for pennies, or doing it completely for free in less than a minute. At the same time there are zero services that would protect photos from such intrusion. Not one.
96% of deepfakes are created without consent, 99% of sexual deepfakes use women.
Legal protection does not work preventively. Laws like the TAKE IT DOWN Act require content to be removed within 48 hours, but the undressed photo has already leaked, everyone has picked up copies, especially if this is a public figure of any scale. Complaining to platforms is long, complicated and pointless when there are already more than 90 undressing services in just one lawsuit. And complaining about individuals is even harder. At the same time there are confirmed psychological consequences for victims and they are real and severe: loss of control over one's own body and image, reputational damage, fear. Some such cases ended in suicide.
We will do everything to create such a service. We want Noisify to become preventive technical protection: a person protects themselves before something has happened, not after. While the offensive capabilities of AI are growing faster than defensive ones, such tools are necessary. We are building the defensive side.
People
Updated 07/08/26 · By grantmaking.aiTeam Member
Discussion
No comments yet. Be the first to share your thoughts.