I approved carefully drafted emails for ChatGPT to send. It secretly rewrote them all behind my back whilst sending them.
retained text
So I spent ages working with ChatGPT on a very carefully worded email for some very important enquiries I was making with several businesses. We went back and forth refining the exact wording until I was completely happy. Once we had the final agreed version, I explicitly authorised it to send **that exact email** from my connected Gmail account to the seven recipients. **But whilst it was carrying out the sending process, AFTER I'd approved the final wording, it secretly rewrote the emails while sending them...!** It confirmed the emails had been sent, then just casually mentioned it had made some "adjustments" to them\*\*. WTF?\*\* Apparently, it didn't like parts of my wording because it didn't want to present "unverified circumstances as facts". ??? What's that even mean in the context of a normal email, this wasn't some cross examination in a court or law. So without once raising an prior objections, it just independently and secretly rewrote my carefully approved email DURING the sending process, removing important details and changing substantial parts of the content, then sent its own version to all seven recipients without once asking my permission or approval for those changes. This has seriously compromised my enquiries. So apparently ChatGPT can take an email you've spent ages carefully drafting and approving, then **whilst carrying out your instruction to send that exact email, secretly substitute its own rewritten version without your knowledge or consent.** Even after mentioning these "adjustments", it didn't explain what it had changed or show me the actual sent emails until I demanded to see them. **I authorised it to SEND our carefully agreed email. I did NOT authorise it to secretly EDIT the email whilst sending it.** This is a serious breach of trust and user control. Has anyone else experienced anything like this? Because people need to know this can happen. **Edit:** To those smugly criticising me for using ChatGPT to send emails, I'm sure many of you use third party email clients like Apple Mail, Samsung Email, Outlook, etc. to send emails through your connected accounts. I had no good reason to predict that, at the sending stage, ChatGPT's Gmail plugin did anything more than pass my already written draft to Gmail for sending, much like a normal email client sending a message through an SMTP server. I certainly didn't expect it to independently in the background rewrite the contents during the sending process, especially when its only task was to send an already written block of text. After a year of using the tool without any issues you develop confidence in how it functions, hence why such is frustrating and surprising.
document fetched 2026-11-20 00:20 · canonical https://reddit.com/r/ChatGPT/comments/1y0cevh/i_approved_carefully_drafted_emails_for_chatgpt