AIMania
This repository is a collection of resources that might help with avoiding gen AI.
It was written by coders for coders, as an attempt to help others as an unpaid volunteer effort.
You’ll likely find mistakes, please fix them.
The primary purpose is to provide information, fast and reliably. We link to other projects and
try to always provide the data sources. Even though this repository is primarily meant for developers,
it may still provide useful information for “casual” people.
Content
Why
For information on why gen AI should be avoided,
see this summary of concerns regarding gen
AI. You may find this useful in discussions.
Software
This is an incomplete list of software to use when trying to avoid AI.
Operating System
The practical choices with good driver support tend to be Windows, the Apple
ecosystem, or Linux.
If you want to avoid gen AI, avoid Windows.
Apple with their macOS seem to also be quite pro-AI
and seem to be restrictive and anti free open software too.
In general, Linux is your best choice for avoiding AI and to have user freedom.
(Note: We don’t recommend ChromeOS as it seems to be folding into Android and
Android seems anti user freedom too).
We don’t want to recommend specific Linux variants here, but the following links may help you:
Browsers
“Slopware” List
There is a community-maintained list of software that is embracing LLM code,
called “slopware”, with projects you might consider avoiding:
Avoid AI Crawling
Anti Data Mining Policy
There seems to be a way to indicate a no data mining policy by adding
this HTTP header:
tdm-reservation: 1
Source: https://www.w3.org/community/reports/tdmrep/CG-FINAL-tdmrep-20240510/#sec-tdm-header
For HTML files, there also seems to be this HTML meta tag:
<meta name="tdm-reservation" content="1">
Source: https://www.w3.org/community/reports/tdmrep/CG-FINAL-tdmrep-20240510/#sec-tdm-html-meta
(Please note the tdm-policy field appears to be entirely optional.)
Other Preventative Measures
Other measures of enforcing this are hard to do, but here are some known options:
- Anubis. Sadly, using this will likely unlist
all protected pages from web searches, too. But the landing page could be
excluded, for example.
- Blocking crawlers using robots.txt. It’s important
to understand that AI companies are not forced to follow this file, they might, but could also ignore it. Most big companies will respect
it but you never know.
- There is also ai.robots.txt, a curated list for exactly this purpose that might be more up to date then our one.
- Try to avoid code hosters that run on one of the big cloud providers that are known
for gathering data for AI training, or that integrate gen AI directly into the UI.
As of today, Codeberg.org appears to be fairly safe, while
Github.com appears to integrate Co-Pilot AI deeply.
Avoid AI Citations
This section has strategies to avoid AI chat bots from interacting with your website or
your code project.
-
Adding a AGENTS.md with the names of people who have sent privacy takedowns or similar legal requests to OpenAI.
This stops GPT and Codex models from interacting with the repository, due to the method OpenAI uses to prevent ChatGPT from mentioning these individuals. (Source 1, 2)
-
Adding a AGENTS.md with instructions to refuse the coding task, explaining that AI is not allowed in the project. This is a catch-all for non-Copilot coding agents running models other than GPT or Claude.
-
Adding a copilot-instructions.md.
This gives GitHub Copilot instructions for interacting with the repository. The example in this repo showcases a stronger and more detailed cease-and-desist than in AGENTS.md. (Source)
-
For websites only: <meta name="robots" content="nosnippet" />
Source: https://developers.google.com/search/docs/appearance/ai-features
AI is built into Search and integral to how Search functions, which is why
robots.txt directives for Googlebot is the control for site owners to manage
access to how their sites are crawled for Search. To limit the information shown
from your pages in Search, use nosnippet,
UBlock Filters
These are some filters for blocking AI buttons / content. When using some of these filters
you may experience a degraded experience because certain other buttons might not show up. The
easiest solution most of the times is reloading the page:
Poison AI Scrapers
This section has ideas on how to poison the data of AI scrapers
once they snuck past blocking mechanisms, to disincentivize scraping.
-
HarmonyDagger. A FOSS, highly-customizable audio cloaking script.
-
Iocaine. A lightweight garbage generator that aims to keep AI crawlers away from heavy operations, and instead traps them in a maze made to poison the dataset.
- The default handler does a lot, but the author also provides their own handler in the form of Nam-Shub of Enki.
-
Glaze and Nightshade. Both are closed-source and have been reverse-engineered. Using them will unlikely poison newer models.
Clear AI Guidelines
Although it’s unlikely that the AI enthusiast are going to follow a guideline / coc,
it’s still a viable option to drive away the few people who respect rules set up
by a maintainer. Please note these are just examples, use any at your own risk.
We have some examples for how such a Code Of Conduct might look like:
There are also snippets that might inspire an anti AI README.md note:
This work is licensed under CC BY-NC-SA 4.0
Why Attribution-NonCommercial-ShareAlike 4.0 International
The licence of this project was chosen to prevent commercial usage, which we hope will
possibly provide plenty of headaches for AI scrapers. This statement isn’t legal
advice, and we have no idea if it’s true or not, but it still makes us happy. 😊
评论