From ebba838680a95129de7581433836157d838c9eb7 Mon Sep 17 00:00:00 2001 From: programmerlexi Date: Thu, 13 Aug 2026 12:56:13 +0200 Subject: [PATCH] Add an entry for the Stack to why_not_llms.md (#1204) ### Description of Changes This patch adds an entry for the Stack v2 dataset to the why_not_llms.md Reviewed-on: https://codeberg.org/ethical-foss/open-slopware/pulls/1204 Reviewed-by: Ethical FOSS admin --- why_not_llms.md | 1 + 1 file changed, 1 insertion(+) diff --git a/why_not_llms.md b/why_not_llms.md index 98ac740..4ea89b5 100644 --- a/why_not_llms.md +++ b/why_not_llms.md @@ -43,6 +43,7 @@ AI companies use data from across the web to train their models, most often with * In 2023, [the Washington Post published a list of sources in Google's C4 data set](https://archive.ph/eehKq). A multitude of fediverse instances and personal sites were included. The fediverse is known for its userbase being major proponents of privacy and opt-in consent, making this especially jarring for those who have chosen to use decentralized social media for control over their data. * In 2025, [a similar leak of Meta's sources was published](https://archive.ph/NZlf3). Meta's list demonstrates how their integration of ActivityPub into their Threads software has enhanced their ability to scrape content without authorization. Threads is widely blocked in some parts of the fediverse, but their scraping of server CDNs has allowed them to get around that. Notably, both the CDN domains of the managed hosting services masto.host and fedi.monster are included in the list; large servers like mastodon.art, which is hosted by the former and has many artists who've left sites like DeviantArt and others due to their AI scraping of user content, had [media unknowingly scraped](https://mastodon.art/@Curator/115022115346692178). * In March 2026, [a research paper](https://arxiv.org/html/2603.20957v2) showed that simply fine-tuning LLMs resulted in outputs containing up to 90% of entire (copyrighted) books, contradicting LLM companies' previous statements in court that their models do not store copies of training data. After fine-tuning exclusively on a single author, the researchers were able to cause the LLM to output works from over 30 completely unrelated authors across different genres. None of the models were explicitly trained on these books by the researchers, which indicates that LLMs always carry with them a considerable amount of copyrighted materials from training. +* In 2026, BigCode released a dataset created from GitHub repositories "with permissive licensing" they scraped in 2025. Searching the dataset using [the provided tool](https://huggingface.co/spaces/HuggingFaceCode/in-the-stack) reveals there are several non-permissive projects in the dataset. The scraping was done without permission and removal request are processed very slow, giving AI companies enough time to download the repositories for which removal was requested. FOSS projects listed in this repo are using tooling that blatantly disregard licensing and violate of Codes of Conduct, making said tools antithetical to FOSS' purpose.