Sam
tante"The texts we choose for our Markov generators should not be random. They should be chosen for their impact on the 'stochastic parrots' that will be trained to speak using our poisonous output. They should strengthen the AI's word-associations between conflicting narratives, where the resolution of that conflict undercuts the system's power."
Luddicus MusThis uses a tool called iocaine, but there are plenty of others that might be suitable, depending on your exact server setup, ethical stance, etc:
https://tldr.nettime.org/@asrg/113867412641585520
ASRG (@[email protected])
Attached: 1 image ## **Sabot in the Age of AI** A list of offensive methods & strategic approaches for facilitating (algorithmic) sabotage, framework disruption, & intentional data poisoning. ### **Selected Tools & Frameworks** - **Nepenthes** — [Endless crawler trap.](https://zadzmo.org/code/nepenthes) - **Babble** — [Standalone LLM crawler tarpit.](https://git.jsbarretto.com/zesterer/babble) - **Markov Tarpit** — [Traps AI bots & feeds them useless data.](https://git.rys.io/libre/markov-tarpit) - **Sarracenia** — [Loops bots into fake pages.](https://github.com/CTAG07/Sarracenia) - **Antlion** — [Express.js middleware for infinite sinkholes.](https://github.com/shsiena/antlion) - **Infinite Slop** — [Garbage web page generator.](https://code.blicky.net/yorhel/infinite-slop) - **Poison the WeLLMs** — [Reverse proxy for LLM confusion.](https://codeberg.org/MikeCoats/poison-the-wellms) - **Marko** — [Dissociated Press CLI/lib.](https://codeberg.org/timmc/marko/) - **django-llm-poison** — [Serves poisoned content to crawlers.](https://github.com/Fingel/django-llm-poison) - **konterfAI** — [Model-poisoner for LLMs.](https://codeberg.org/konterfai/konterfai) - **Quixotic** — [Static site LLM confuser.](https://marcusb.org/hacks/quixotic.html) - **toxicAInt** — [Replaces text with slop.](https://github.com/portasynthinca3/toxicaint) - **Iocaine** — [Defense against unwanted scrapers.](https://iocaine.madhouse-project.org) - **Caddy Defender** — [Blocks bots & pollutes training data.](https://defender.jasoncameron.dev) - **GzipChunk** — [Inserts compressed junk into live gzip streams.](https://github.com/gw1urf/gzipchunk) - **Chunchunmaru** — [Go-based web scraper tarpit.](https://github.com/BrandenStoberReal/Chunchunmaru) - **IED** — [ZIP bombs for web scrapers.](https://github.com/NateChoe1/ied) - **FakeJPEG** — [Endless fake JPEGs.](https://github.com/gw1urf/fakejpeg) - **Pyison** — [AI crawler tarpit.](https://github.com/JonasLong/Pyison) - **HalluciGen** — [WP plugin that scrambles content.](https://codeberg.org/emergentdigitalmedia/HalluciGen) - **Spigot** — [Hierarchical Markov page generator.](https://github.com/gw1urf/spigot) --- *This is a living resource — regularly updated to reflect the shifting terrain of collective techno-disobedience and algorithmic Luddism.*
Pol@tante while the idea is noble and good, I would assume that such text generation is a waste of time. Companies that train new/bigger LLMs can just remove the poison with existing LLMs.
Dieu@hllizi @tante they are already scraping ai slop from millions of scammer pages and ai slop blogs? They must know that and must already be working hard on avoiding that content. If you want to defend your own content, look at captchas, anti-scraping cdn, make sure there is no content when javascript is not executed, check if images are loaded, implement homegrown proof-of-work challenges etc.