I have been making art for 26 years. I stopped posting my finished work on open platforms years ago because I watched what happened to it. Bots scrape it. Print-on-demand suppliers upload it to Temu and AliExpress. AI companies train on it. The image I spent weeks making becomes a file someone else profits from, and I find out about it months later if I find out at all.
So I went looking for every protection tool that exists. I wanted to know what was actually out there before I built anything myself. Here is what I found — and here is my honest assessment of every single one of them.
Glaze and NightShade — the most popular tools, and they are already broken
Glaze and NightShade are the most widely used image protection tools for artists. Together they have been downloaded almost nine million times. They were developed by researchers at the University of Chicago and are free to use. Both work by adding invisible distortions — called poisoning perturbations — to your image before you upload it. The idea is that these distortions confuse AI models during training so they cannot accurately learn your style.
Glaze takes a passive approach, hindering the AI model's ability to extract your stylistic features. NightShade goes further, actively corrupting the learning process by causing the AI to associate your style with completely unrelated concepts.
In June 2025, a team of researchers at the University of Cambridge and the Technical University of Darmstadt published a paper at the USENIX Security Symposium. They had developed a method called LightShed that can detect, reverse-engineer, and remove both Glaze and NightShade's protections — with up to 99.98% accuracy. The invisible distortions are stripped away and the image is restored to its original, fully usable state for AI training.
Nine million downloads. Broken by a single research team in one paper. Artists who have been relying on these tools are still at risk and most do not know it yet.
Watermarks — removed in one click
Watermarks are the oldest and most common image protection advice for artists. Add your name, your website, your copyright notice over the image. Make it hard to use without your involvement being obvious.
AI-powered inpainting tools — now built into free consumer applications including Adobe Photoshop, GIMP with plugins, and standalone tools — can identify a watermark, remove it, and reconstruct the image underneath in seconds. What took hours of careful manual editing five years ago now takes one click. Watermarks deter the casual, unsophisticated thief. They stop no one with actual tools and intent.
There is also the specific case of artists whose signature is part of the work itself. My own signature is part of how I paint. Temu suppliers selling Australian artist Lottie Rae's work did not even bother to crop her signature out of the image. It is still there, in the corner of the product listing, as her work sells for AU$5.42. Her name in the corner did not stop a single sale. It was simply irrelevant to the automated process that took her work.
Right-click disabling — stops no bot, ever
Right-click disabling prevents the browser context menu from appearing when someone right-clicks on your image, removing the "Save image as" option. It is so trivially bypassed that listing the bypasses is almost embarrassing: keyboard shortcuts, drag and drop into a folder, browser developer tools, screenshots. And automated scraping bots do not right-click at all. They read page source code, find image URLs, and download directly from the server. The browser interface is never involved. Right-click disabling provides zero protection against the automated systems that supply Temu.
Low-resolution uploads — upscaled in seconds
The advice here is to upload images at a resolution too low to be useful for printing. If the file is only 600 pixels wide, it cannot be printed on a large canvas without looking terrible.
AI upscaling has made this advice obsolete. Free tools — including Topaz Gigapixel, Real-ESRGAN, and tools built into consumer applications — can take a low-resolution image and generate a higher-resolution version with significant detail reconstruction. The upscaled quality is not perfect but it is more than sufficient for the sizes and print qualities Temu and similar platforms use. A 600-pixel image that gets upscaled to 3000 pixels is printable at standard poster and canvas sizes.
Low-resolution uploads also directly harm your legitimate sales. A collector who cannot see your work properly at full quality on screen is less likely to buy. The protection hurts your business while providing only limited deterrence.
Kudurru — reactive bot blocking
Kudurru is a plugin that detects scraping bots and adds their IP addresses to a block list, preventing them from accessing your site. It can also send corrupted or deceptive images back to identified bots — poisoning the data they collect.
The limitation is that it is reactive. A bot that has not yet been identified scrapes your site before it is blocked. Sophisticated scrapers rotate IP addresses specifically to avoid block lists. And Kudurru requires technical setup and ongoing management that most artists are not equipped to do. It is a useful layer of defence for technically capable artists who already have their own websites. It is not a structural solution.
Robots.txt and crawler blocking — only works on compliant crawlers
The robots.txt file is a standard web file that tells crawlers which pages they are allowed to access. Many AI companies now say they respect opt-out mechanisms in robots.txt. Some website hosting platforms allow you to specifically block AI crawler user agents.
The problem is compliance. Legitimate companies like Google and Bing honour robots.txt because they have a commercial incentive to maintain trust with website owners. Scraping bots that supply Temu suppliers do not have that incentive. They are not running as identifiable user agents. They do not care about robots.txt. The mechanism works for crawlers that choose to respect it. The crawlers doing the most damage choose not to.
Timestamp certification and blockchain — proves ownership, does not prevent theft
Platforms like ArtMajeur offer certified timestamps when you upload work — a record that proves you uploaded this image on this date. Some services use blockchain to create an immutable provenance record.
These tools are valuable for proving ownership after theft has occurred. They do not prevent the theft from happening. A certified timestamp tells a court you created the work. It does not stop a Temu supplier from uploading it to a print-on-demand service the day after you post it.
Bluekea's protection mask — the closest thing to a structural solution
Bluekea is a photography platform that offers a protection mask system. A high-opacity layer covers most of the image, with only a small sharp portion visible at any one time. The sharp region moves as the viewer moves their mouse or finger across the image. The viewer can explore the whole photograph but the complete image is never fully visible on screen simultaneously.
This is the closest thing I found to a genuinely structural approach anywhere in the market. The concept is sound — if the complete image never exists on screen at once, there is no complete file to capture. The implementation is limited to a photography portfolio platform, not an art sales marketplace, and it has not been built with the depth and specificity needed for fine art display.
What I built instead
After going through all of this, I built The Grid for Solene Haus — the platform I founded for Australian artists.
The Grid operates on the same structural principle as Bluekea's mask but built specifically for fine art display and extended significantly further. Every artwork on Solene Haus is displayed through a tile system where the complete image never loads on screen at any moment. At rest, every tile is heavily blurred — not softly, not artistically, but substantially enough to destroy usable detail while leaving the colour, composition, and mood of the work fully visible. When a collector hovers or taps, exactly one tile sharpens while every other tile remains blurred. The sharp tile follows the collector's movement across the canvas, allowing them to explore every inch of the work in full detail.
A scraping bot visits the page and finds only blurred tiles. A screenshot captures only the blurred state or a single sharp fragment. The complete image URL does not exist in the page source because the complete image never loads as a single file. There is no complete image to steal because the complete image never exists on screen.
A collector can fall completely in love with the work — they can explore every brushstroke, every colour decision, every detail — without the complete file ever being available to take.
No watermark to remove. No resolution to upscale. No perturbation to strip. No block list to rotate around. The stealable object simply does not exist.
The honest summary
Every protection tool that existed before structural display prevention was built around one of two approaches: making the image harder to use after it has loaded, or making the image less useful before it is uploaded. Both approaches share the same flaw — the complete image exists somewhere, either on screen or in a file, and anything that exists can be obtained by someone sufficiently motivated.
Structural prevention removes that premise. The Grid means the complete image never exists on screen. That is not a deterrent. It is a different architecture.
If you want to understand exactly how The Grid works and see it in action, it is live on every artwork displayed at Solene Haus. I built it because nothing else was doing what actually needed to be done.