Tools to Disrupt the Unethical Scraping of Art: My thoughts on Glaze and Nightshade

 For my anti-scraping tools experiments, I chose a personally meaningful AI-generated image from 2018. Created using an early version of Google's Deep Dream Generator in style transfer mode, the image merges a photograph of my right eye (taken during a rejection episode of my then three-decade-old corneal graft) with a cloud formation.

The creation process involved using the cloud image to establish the training set and then applying this "style" to a photograph of the right side of my face, originally intended as a profile picture. While I was pleased with the full-sized result, the cloud effect became less visible when reduced to smaller dimensions, making it appear simply as an unusual eye image. Though I ultimately didn't use it for its original purpose, this unique composition proved ideal for testing anti-scraping technologies.

This image represents an interesting circular journey—AI-generated art now protected from newer AI generative systems—making it particularly appropriate for this blog post.

Two emerging technologies from research indirectly supported at the University of Chicago offer some hope: Glaze and Nightshade. Though both tools share somewhat technical interfaces and require significant processing time on standard computers, they serve different protective functions:

Glaze embeds inappropriate AI training signals into your images, making it difficult for generative AI to accurately reproduce your artistic style when prompted by your name. My test on an eye/cloud composite image (see above) took approximately three hours on a standard i5 computer. The resulting image appears visually similar to humans but contains embedded neural network weights that should confuse any AI machine learning systems. While this doesn't prevent scraping, it reduces the likelihood of your style being accurately replicated. Which makes "glazing" suited to any artist seeking to protect their unique look.

Nightshade takes a different approach by "poisoning" specific text prompts associated with your work. This tool processes faster (about 30 minutes in my test) and creates images that appear normal to humans but contain corrupted data that could potentially damage AI training datasets. The theory is that widespread adoption might eventually reduce the reliability of AI models trained on scraped content, though this remains speculative. I'm currently testing Nightshade on some of my online photos to gauge any noticeable effects or responses.

Both tools add identifiable suffixes to filenames that might eventually be recognized by scrapers, ironically helping them avoid these protected images. This might in part be what artists want, those internet giants not using or freely sharing our work. Encouraging them to contact us, seek permission and pay to use would of course be better. Additionally, social media platforms typically strip metadata and rename files upon upload, potentially limiting the effectiveness of this possibility.

For maximum protection, artists can apply both techniques sequentially, though the practical impact of this combined approach is still unclear. 

While these technologies represent steps toward better intellectual property protection, they're early solutions in what will likely be an ongoing technological and legal conversation about artists' rights in the AI era.

Tools to Disrupt Unethical Scraping of Art: My Experience with Pixsy

Several years ago, I beta-tested through flickr an application now known as Pixsy.com that scans the internet for unauthorized uses of protected images. I continue to use the free version, which offers limited scanning capabilities, though paid tiers provide expanded monitoring and response options.

The system effectively allows users to review potential matches, verify ownership, and decide whether to ignore usage, issue takedown notices, or pursue other legal remedies. Since most of my Flickr images are posted under Creative Commons licenses permitting non-commercial reuse with attribution, I typically verify whether users have followed these terms.

While I've received occasional permission requests (which I usually approve for non-commercial users), I've only needed to issue one takedown notice when a local government publication used my work without acknowledgement. Though the digital version was removed, I suspect printed copies had already been distributed. Given the circumstances and potential legal costs, I chose not to pursue further action.

My recent monthly summary showed a significant increase in matches from social media platforms, particularly:

- Google, despite no longer actively posting to Google Photos (my content should be private except for occasional Blogger posts)

- TikTok matches via a "TikTok scraper" application (I have no TikTok account or uploads)

- Instagram matches (while my content is viewable there, it shouldn't be downloadable)

This unexpected proliferation of my images across platforms I believed were private or don't use, highlights the ongoing challenges in protecting digital creative work. Monitoring tools like Pixsy are a reasonable start. Still, there is obviously more to be resolved to ensure creatives don't have their work plundered to make a profit for unscrupulous others without permission or compensation.

A little bit about "Littles"

 I watched an interesting YouTube demonstration by Al Klein on the American Art School Live Channel about a little tool/technique he uses for plein air sketching. He creates what he calls “littles”. These are small colour studies/composition sketches/almost tonal studies. He creates them in his sketchbook and uses a small viewfinder frame that he made to be the exact ratio of his potential larger work later. 

His beautifully simple next step is to place the viewfinder on a blank sketchbook page and use the cutout area as a template and trace out the frame shape onto the page. Thus eliminating any virtual scaling in your brain when transferring your sketch to a larger studio work. It also helps firmly position the main shapes you observe into the correct position within the composition.

I don’t necessarily agree with all his terminology and explanation, for example, colour temperature is independent of tone. I prefer to limit the term tone to describe the difference between light and darkness. what I did like is the idea of a small sketch in the field that looks at seeing and defining those shapes of colour and tone. So I had to make up a couple of viewfinder frames that would suit any A5 sketchbook.  I like working initially in an A5 sketchbook on site with some basic tools, a small portable colour palette and a pencil. I have made up a variety of such kits based on cheaper A5 visual diaries/sketchbooks to do my little experiments in the field. I’ve also started using them pretty exclusively for my tonal studies in the style of Andy Evanson, these are just monochrome and all about understanding where to put the initial washers and how to join the various toned shapes. They definitely are not colour studies.

I happen to have some internal offcuts of Mat from frames and set about cutting my viewfinder frames. I’m not using sheet and half sheet sizes for my watercolours much anymore I tend to work in sizes A3, A4 and A5. So decided to try out some viewfinder cutouts based on smaller A-size ratios. I quickly dismissed the A6 size hole as it was too large, my arm wasn’t long enough. Basically, when you hold it in front of your eye it takes in too much of the view to be useful in terms of refining a decent composition. So I ended up cutting 2 frames, one with an A7 size hole and one with an A8 size hole. The A7 size is probably my favourite and works very well at roughly arm's length, you can swing it around and see the various compositions very quickly. The A7 viewfinder seems incredibly small but it’s very useful for zooming in and out, moving it closer or further from your eye and can make an outline for a very neat little colour study in your sketchbook.

The issue of finding the best framing is not just limited to plein-air artwork, it also applies to photography. The vast majority of folk still just lift the camera to their eye roughly centred on the subject they are photographing and press the shutter. A small amount of time spent changing the view (frame) can have remarkable results. Changing a ho-hum composition into something striking. This is really simple with a modern digital camera if they have a suitable LCD screen on the back to preview your test photograph and is really straightforward on any modern smartphone with a camera. Remember to try the zoom feature to get closer or a broader view. If you don’t have a zoom lens you can always zoom with your feet.

Why I've been missing.

 I'm not one for excuses but I don't mind sharing what has been slowing me down on the computer/internet/socials front.

1. My recent corneal graft is rejecting and I am losing my vision in my right eye. This makes any work on the computer very taxing.

2. Discovered that some of my best work has been "scrapped" (without permission) into a couple of the more significant Large Language Model data training sets.

I view this as unethical and dangerous behaviour by those driving the current Race to dominate the so-called "AI race". By the way, the current crop of generative and large language tools are not intelligent. They just appear superficially intelligent.

 My current solution, which isn't so good, is to stop publishing my original works on the net.

3.  My reach on Instagram has dramatically declined and I can't be bothered anymore to be part of the constant scramble to keep a users attention so some people can get very rich from advertisers.

Stop the death scrolling folks get up and enjoy life!

You can see more of  adansito work at NightCafe