Guard

An AI that reads privacy policies for you

5.0ā€¢1 reviewā€¢

18 followers

Visit website
ā€¢
Did you know Tinder shares your private conversations and matches? They're pretty straightforward about it in their privacy policy. Guard is an Artificial Intelligence that reads privacy policies for you and spots the threats in them. Protect your privacy.
Guard gallery image
Guard gallery image
Guard gallery image
Guard gallery image
Guard gallery image
Guard gallery image
Guard gallery image
Guard gallery image
Guard gallery image
Launch tags:Privacyā€¢Artificial Intelligence
Launch Team
AssemblyAI
AssemblyAI
Build voice AI apps with a single API
Promoted

What do you think? ā€¦

rameerez
Maker
Hi PH! šŸ‘‹ So excited to be launching something again! šŸš€ Meet Guard, an šŸ§  Artificial Intelligence that reads privacy policies for you and warns you of the biggest privacy threats in the digital services you use. šŸ¤“ This is actually the basis of my thesis on AI and Natural Language Processing (NLP), and it's been super fun to work on this problem with such an interesting technology. For the moment Guard is a website that spits out some of the AI results in a simple, organized and visual way. It also includes some other relevant information such as recent data breaches and instructions to delete your accounts and data on those services. But I'd love to hear your feedback on what should this project become next! šŸ’» Actually, I've been working on a spinoff: an app that helps you protect your digital privacy. It's like an antivirus, but for privacy threats instead of viruses. This is only a rough prototype right now, but I'd also love to hear your thoughts on it! https://useguard.com/blog/future/ Also happy to answer any Qs about the tech behind this! :)
Shivendra Soni
@rameerez While opening on my corporate network. I get error like it has malware HTML-genA
Ludwig
Wow, congratulation on the launch, inspiring product! Curious on how the scores adapt if the privacy policy changes? Will it then be scored based on the trained AI - which at best is already generalised enough?
rameerez
Maker
@ludwig_stumpp Thank you! These are my plans for when a privacy policy changes: Guard should automatically fetch it, re-analyze it and re-calculate the score based on the new input. This should be a rather simple process, provided the AI already knows how to analyze new unseen input. Measuring under/overfitting and aiming for correct generalization is something I'm doing for the thesis so hopefully this automatic process will work with good precision!
Pete
nocsdegree
Looks great! I think you should add the fact about Tinder on here so people are aware.
rameerez
Maker
@petecodes Good tip, just added! Thank you Pete!
Luke Oslizlo
CleanShot
I love the idea, because letā€™s be honest - nobody has time to read the super long privacy policies
rameerez
Maker
@lukaszmtw Thanks! Interesting fact -- from research I've read, this number is actually measured: only 0.001% of internet users even **start** reading privacy policies, so the number of people that **finish** reading them must be much lower! And, if you had to read all the privacy policies you accepted just in the past 5 years, you would need 3.040h of non-stop reading!
Akshay Kadam(A2K)
Amazing product. What kind of tech did you use for Guard?
rameerez
Maker
@deadcoder0904 broadly speaking, it's deep learning: it's a deep neural network (based on a specific type of recurrent neural network architecture) that models the language and "understands" what it says to classify it either as friendly or dangerous. I can also get into the super technical details if you'd like :) Regarding the actual tools I used, it's Python 3.6 + Pytorch for the AI part (and a bunch of the typical libraries like pandas, nltk, numpy et al.), and Ruby on Rails for the web app.
Alex Wawl
@deadcoder0904 @rameerez " I can also get into the super technical details if you'd like :)" - I would love to here about technical details, I'm really interesting about it. One little question by the way - where did you get datasets for training? In the internet you can find a lot of "standard" datasets which help to detect spam/bulling/sentiment analyze/etc but how do you define "bad" and "good" words for privacy?
rameerez
Maker
Liam Roberts
Super interesting product! Keen to see how it develops. Can you please link us to your own Privacy Policy? I cannot seem to find it on your website. My suggestion would be to have your Privacy Policy a touch more accessible especially if 1. you are a company criticising poor Privacy Policies; and 2. you are generating data from users with the 'Teach the AI' function.
rameerez
Maker
@liam_roberts Sure, it's here: https://useguard.com/blog/privac... There's more info about it in /experiment Also, this is mostly a personal project (vs a company) + users are informed how exactly their data will be used (see /experiment) + this is mostly for science (did this project as the basis of my academic thesis)
Liam Roberts
@rameerez, thank you!
Daniel Teo
Storyboarder
Great job! Possible to make it into a chrome extension and flag to the user should the privacy policy on a website have a very poor grade?
1234
Next
Last