🧵 Anti-AI folks: I tried out Read.AI so you don't have to. DO NOT allow this AI on your system for any reason. I have talking points from an AI industry expert for anyone arguing against implementation of Read.AI (a part of Copilot) at their institutions:
Thank you!
Correction: This isn't a part of Copilot officially, but Microsoft has been pushing it very heavily via 365 and Teams. That's how I got this crap on my system: It was pushed via Outlook.
1. Outlook: Read.AI doesn't understand listservs, spam, and mass emails. It assumes that these are meant for you personally. If someone sends a meeting request via email, and you don't have anything on your calendar during that time, it will sign you up without your knowledge or consent.
2. Outlook: Read.AI again assumes that all email coming into your inbox is relevant to you. Its summaries use the second person like a boss who keeps changing your job description. "You're working in this project with this college, and here are your next steps" when it was a bulk email.
3. Outlook: Read.AI does not understand working hours. To make its calendar function remotely useful, you would have to block out all times in which you're NOT working. Like 2 a.m. on a Sunday. Or else it will tell everyone who emails you that you'll be at their meeting.
4. Zoom/Teams: Read.AI promises to take notes during your Teams and Zoom meetings. I quite like Zoom's AI notetaker, and the two do not get along. I also like human notetakers. Read.AI's meeting transcription spams everyone in the meeting whether they're signed up for Read.AI or not.
Found something deep in the documentation: To remove from Zoom, go to marketplace.zoom.us/user/installed when logged into Zoom, and remove Read.AI.
When it sent me the email that I'd removed it from Zoom, it also had a link to delete my account. I never created a Read.AI account, so first I had to sign up for one, and then delete it. Hopefully this will nuke it from my Outlook.
5. Privacy laws: Read.AI's invasive approach pings my FERPA/COPPA/HIPAA compliance radar something awful. Imagine you're a Title IX advocate (as I am) and you get an emailed summary of a complaint from a student. Read.AI digests that email. It won't anonymize it. This is a privacy disaster.
6. Removal/Opt-Out: I have found instructions from Microsoft about how to remove Read.AI permissions, and all of them are inaccurate. The menus don't exist, or Read.AI doesn't show up on Microsoft 365 like it should, or I need admin access (which many school/healthcare employees don't have).
Takeaway: It's fun for me to dunk on comically bad AI like Google Gemini. But Read.AI is a downright sinister piece of work. It's badly programmed with no user experience (UX) in mind. It's invasive. It violates privacy laws. And it's almost impossible to remove. Institutions need to block it. Now.
tag @unroll.bsky.social
Thank you for your valiant service! Not even halfway exaggerating!