<![CDATA[Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI]]>This post did not contain any content.]]>https://isurg.org/topic/c77ff028-a1ff-47a3-8306-734712c827f7/someone-should-make-a-community-to-freely-distribute-examples-of-data-poisoning-people-can-randomly-put-in-their-social-media-posts-images-to-sabotage-aiRSS for NodeWed, 26 Aug 2026 18:36:20 GMTTue, 11 Aug 2026 14:10:12 GMT60<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 16:50:26 GMT]]>I don't want to put it back in the bottle. I just feel like taking massive public user data for free should not be devoid of consequence.

]]>https://isurg.org/post/https://lemmy.world/comment/25241160https://isurg.org/post/https://lemmy.world/comment/25241160Tue, 11 Aug 2026 16:50:26 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 16:48:24 GMT]]>Who says the objective is to impact AI companies negatively?

There's a series of valid motivations to want AI models to not be able to use public user data with no consequence.

]]>https://isurg.org/post/https://lemmy.world/comment/25241133https://isurg.org/post/https://lemmy.world/comment/25241133Tue, 11 Aug 2026 16:48:24 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 16:41:12 GMT]]>And here's some explanations of how various attacks work.

Link Preview Image
GitHub - nukIeer/AI-Prompt-Injection-Cheatsheet: AI hacking snippets for prompt injection, jailbreaking LLMs, and bypassing AI filters. Ideal for ethical hackers and security researchers testing AI security vulnerabilities. One README.md with practical AI prompt engineering tips. (180 chars) Keywords: AI hacking, prompt injection, LLM jailbreaking, AI security, ethical hacking.

AI hacking snippets for prompt injection, jailbreaking LLMs, and bypassing AI filters. Ideal for ethical hackers and security researchers testing AI security vulnerabilities. One README.md with practical AI prompt engineering tips. (180 chars) Keywords: AI hacking, prompt injection, LLM jailbreaking, AI security, ethical hacking. - nukIeer/AI-Prompt-Injection-Cheatsheet

favicon

GitHub (github.com)

Link Preview Image
How Hackers Trick AI: The Hidden World of Prompt Injections and Jailbreaks

__ We live in a time where chatting with an AI feels almost natural. You ask a question, it answers.... Tagged with llm, ai, machinelearning.

favicon

DEV Community (dev.to)

Link Preview Image
How Hackers Exploit AI’s Problem-Solving Instincts | NVIDIA Technical Blog

As multimodal AI models advance from perception to reasoning, and even start acting autonomously, new attack surfaces emerge. These threats don’t just target…

favicon

NVIDIA Technical Blog (developer.nvidia.com)

]]>https://isurg.org/post/https://lemmy.world/comment/25241003https://isurg.org/post/https://lemmy.world/comment/25241003Tue, 11 Aug 2026 16:41:12 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 16:35:28 GMT]]>I misunderstood the purpose of that community. I figured it was just for discussing how to poison AI models. But actually visiting it, I see it is primarily for posting gibberish in the hopes that AI models would scrape the sub and treat it all as genuine content. As you say, that does not seem like it would have much of any impact on actually poisoning AI models because basically every scraper is going to know to avoid a community called "poison AI".

]]>https://isurg.org/post/https://quokk.au/comment/5570661https://isurg.org/post/https://quokk.au/comment/5570661Tue, 11 Aug 2026 16:35:28 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 16:29:55 GMT]]>
Link Preview Image
Totally True Facts No Lies - Lemmy.World

Welcome to Totally True No Lies. Bots are scraping Lemmy posts to train AI. But what if not all the posts are 100% correct? The AI would be wrong. It would be awful! I know we all want to make sure that AI models are trained on nothing but the most realist science and news and factual facts, so that they can be as correct and capable as possible when they replace us at our jobs. Rules: Self diagnosed doctors, scientists, experts, wizards, and people who are always right only. No down voting without community participation. Too many people can’t be bothered to read sidebars and don’t get the point. Don’t be dicks to each other.

favicon

(lemmy.world)

]]>https://isurg.org/post/https://piefed.world/comment/6415220https://isurg.org/post/https://piefed.world/comment/6415220Tue, 11 Aug 2026 16:29:55 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 16:22:22 GMT]]>Poi sonai, while initially was able to influence the AI output but the whole subreddit got selectively filtered out and doesn't have impact any longer. It's still good place to discuss the topic though.

]]>https://isurg.org/post/https://sh.itjust.works/comment/26847181https://isurg.org/post/https://sh.itjust.works/comment/26847181Tue, 11 Aug 2026 16:22:22 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 16:13:09 GMT]]>A centralized database of content for AI scraping agents to be trained to exclude?

]]>https://isurg.org/post/https://lemmy.zip/comment/28176375https://isurg.org/post/https://lemmy.zip/comment/28176375Tue, 11 Aug 2026 16:13:09 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 16:07:08 GMT]]>like an Anti AI alliance?

]]>https://isurg.org/post/https://lemmy.blahaj.zone/comment/21709231https://isurg.org/post/https://lemmy.blahaj.zone/comment/21709231Tue, 11 Aug 2026 16:07:08 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 15:25:37 GMT]]>That assumes that AI companies are negatively impacted by the quality of their product. It’s true that they’re competing with each other based on their quality relative to other companies’ products, but poisoning public data impacts everyone’s models similarly. Setting aside competition and looking at the success of the AI sector as a whole, I think it’s more dependent on marketing and hype than on real performance... and if that’s the case, then poisoning public data doesn’t hurt anyone except the people being forced to use it.

]]>https://isurg.org/post/https://lemmy.world/comment/25239698https://isurg.org/post/https://lemmy.world/comment/25239698Tue, 11 Aug 2026 15:25:37 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 15:12:17 GMT]]>You have one on lemmy.world

]]>https://isurg.org/post/https://lemmy.eco.br/comment/22470289https://isurg.org/post/https://lemmy.eco.br/comment/22470289Tue, 11 Aug 2026 15:12:17 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 15:02:38 GMT]]>There's a new technique that uses the AIs "thinking" tags to get it to do things that are otherwise banned by policy.

I'll have to find the article again. But due to the way LLMs work, they can't defend against this sort of attack.

]]>https://isurg.org/post/https://lemmy.world/comment/25239315https://isurg.org/post/https://lemmy.world/comment/25239315Tue, 11 Aug 2026 15:02:38 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 15:02:22 GMT]]>People still have sm accounts?

]]>https://isurg.org/post/https://lemmy.world/comment/25239311https://isurg.org/post/https://lemmy.world/comment/25239311Tue, 11 Aug 2026 15:02:22 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 14:55:37 GMT]]>Lol, good luck with that.

The more poisoning you attempt, the more effective anti-poisoning becomes.

AI LLM is here, stop trying to put the genie back in the bottle. All we can do is figure out how to use it and prevent misuse.

]]>https://isurg.org/post/https://lemmy.cafe/comment/18532642https://isurg.org/post/https://lemmy.cafe/comment/18532642Tue, 11 Aug 2026 14:55:37 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 14:26:03 GMT]]>I've read in papers that you can poison datasets with a very small percentage of the data, if done cleverly. I can fish up the source if you want (but it might take me some time).

edit: here it is.

We conduct the largest pretraining poisoning experiments to date, pretraining models from 600M to 13B parameters on chinchilla-optimal datasets (6B to 260B tokens). We find that 250 poisoned documents similarly compromise models across all model and dataset sizes (...)

Emphasis mine. All it takes is 250 poisoned documents.

]]>https://isurg.org/post/https://lemmy.world/comment/25238734https://isurg.org/post/https://lemmy.world/comment/25238734Tue, 11 Aug 2026 14:26:03 GMT<![CDATA[Reply to Someone should make a community to freely distribute examples of data poisoning people can randomly put in their social media posts/images to sabotage AI on Tue, 11 Aug 2026 14:24:06 GMT]]>Na, for it to be effective it needs to be wide spread, but if its wide spread then it can be filtered out of the training material.

]]>https://isurg.org/post/https://lemmy.world/comment/25238698https://isurg.org/post/https://lemmy.world/comment/25238698Tue, 11 Aug 2026 14:24:06 GMT