Hiya, just quickly wondering if anyone know about a good tool for comparing Privacy policies against each other? Im currently downloading each PP, then using self-hosted StirlingPDF to compare 1 on 1. However, I am looking for a more efficient tool, to compare multiple at the time, if there are any. Any tool that can handle multiple PDFs or HTML files and look at the differences between them kinda tool.
Appreciate any suggestions! 🕵️
Privacy has become a very important issue in modern society, with companies and governments constantly abusing their power, more and more people are waking up to the importance of digital privacy.
In this community everyone is welcome to post links and discuss topics related to privacy.
[Matrix/Element]Dead
much thanks to @gary_host_laptop for the logo design :)
There is a general rule, the longer and more claused a TOS/PP is, the more they intend to hide. TOS and PP are legal documents and companies are required by law to specify their policy, if they track and profile the user they must specify it in these documents, for this reason they try to use long texts with a lot of legal jargon as much as possible, so that the user gets bored of reading it completely and does not understand even half of it. Tosdr is fine, but lacks still the analysis of much sites, now with the proliferation of AI services even further. Only very few have a TOS/PP like eg Andisearch, most others are welcome tools used by large corporations to collect user data.
https://www.zzzuckerberg.com
Hahaha haha I cannot believe someone made https://www.zzzuckerberg.com, that’s just freaking hilarious! Love that the brightness also decrease the further you scroll 😂
I think there’s a privacy TLDR thing.
Are you looking for a tool that can diff legal documents line by line or clause by clause? If the latter I’d bet an LLM with a large context size could do a pretty good job, especially if you used a script (or another pass through the LLM) to break them down into like sections so that could just compare e.g. all Controlling Law sections with each other and all IP Indemnification sections with each other.
Now that I think about it, tuning the prompt (and keeping the temperature very low, like 0) you could probably get it to return everything from proper diffs to summaries of conceptual differences. And it could definitely do multiples at once if you were to break them into like pieces ahead of time.
Preferably line by line. Kind of like what Github does whenever you apply a commit, it will make a red line for what is removed and a green line for what is added code. I could look into LLMs though, but was hoping to find a quick n dirty tool to do the job.
After reading this, I’m thinking whether converting the PDFs to markdown and diffing them with a text difftool could work.
If you go this route, you may want to test with different diff algorithms. Git has multiple too, but I don’t remember right now which I found to be the best
Thanks for the tip!
Now that I’m at my computer, I was able to find the diff alg I was thinking about: it’s
histogram
.Here’s an issue from gitea about when they changed the default git diff alg to this one: https://github.com/go-gitea/gitea/issues/23255
And here’s an article I have found earlier about some of the available git diff algorithms, and when they are too be used: https://luppeng.wordpress.com/2020/10/10/when-to-use-each-of-the-git-diff-algorithms/
thanks very much for sharing ☘️
Like a diff checker?
https://www.diffchecker.com/
This is pretty close to what im looking for actually, thanks for sharing! :)
Glad to help!
Tosdr
While this is a great service, it’s not a tool that allows me to enter my own downloaded privacy policies and compare against others, as far as I’ve understood it? This seems like a service that you can upload a PP too and wait for it to be processed.
I feel like that’s better, as long as it is a human doing the processing, rather than a program algorithmically output a closest case answer.
I’m only going through 10 of them, so personally it would be quicker to do it manually.