43
Tools we can use to poison AI and fight back
(lemmy.today)
Just talk nonsense all the time
Just fyi, LightShed was released over a year ago and can strip poisons from images. Glaze has become partially ineffective and Nightshade is broken.
To be clear, it's still better to use even a partially functioning poisoning tool than none at all.
LightShed isn't perfect, though, and introduces its own AI bias to images, basically creating AI incest. Plus, it takes them time to run it. Poisoning is still a good idea. Although, yes, LightShed is a problem.
FWIW, while iocaine does generate copious amounts of garbage, it is never going to poison LLM training data: it's trivial to filter iocaine-generated garbage out pre-training. The garbage is there for two reasons:
- To serve the crawlers something.
- To poison the crawling queue.
Not the models, the crawling queue. Place an identifiable substring into every generated URL, and you can catch any crawler that works off of a queue collected by cheaper crawlers.
It's not really a poisoning tool - it originally started with that goal, but that goal has changed a loooong time ago -, it's more bot detection & mitigation nowadays, the garbage is just one of the tools it achieves this goal with.
But while I'm here, I'd like to mention Alun Jones' fakejpeg (or my rust port): a library that can help you generate structurally legit-ish JPEGs, files that most decoders will happily display, but with entirely random content. Generating them is very cheap. How effective it is in poisoning? I don't know. But I find great delight in looking at the random noise.
Fauxx is pretty cool.
Can you provide a link, either via comment or by editing the text in your original post?
EDIT: I found it.
As interesting as it is, it has a few problems.
LLM use in the project
LLMS, as we all know, have a vast number of problems. That's why we're in this community.
The biggest issue with Fauxx is that it uses LLMs for documentation, issues, and in "assistance" writing code and fixing bugs. Any programmer who outsources all or part of their thought or work to a slop bot is not exactly smiled upon in my view or here on !fuck_ai@lemmy.world. See the project's gen-AI use disclosure.
I believe that these two statements in that disclosure are mutually exclusive:
-
No, this project was not vibecoded.
-
I also use it [generative AI] for assistance in writing code or chasing down bugs.
Hosting on GitHub
The second most notable issue to me is that it is hosted on GitHub, which is less than ideal. One of the reasons it is not ideal is because all of its source code has, by virtue of its hosting on GitHub, been scraped by Micro$lop Copilot.
If this is the level of your moral purity tests, I would suggest you just stop using the internet at this point.
this post was submitted on 29 Oct 2026
43 points (95.7% liked)
Fuck AI
8584 readers
1447 users here now
"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
AI, in this case, refers to LLMs, GPT technology, and anything listed as "AI" meant to increase market valuations.
Rules:
- Relevant posts only.
- No bigotry.
- This isn't a place to talk about the cons and (supposed) pros of AI for different use cases. This is /c/fuck_ai .
- Especially no AI advocacy.
- Don't make any posts that are just AI content with no context unless the point of your post is obvious from content. (Don't just post slop advertisements or slop political cartoons. At least make your point about it clear so we know you aren't just trolling.)
- Venting is part of the purpose of this community, but don't aim frustration or accusations at other users.
- These rules are a living document that may change in the future.
founded 2 years ago
MODERATORS