ChatGPT Google Drive Safety Blocks — Engineering Review Requested

ChatGPT Bugs
1

Environment: ChatGPT web, Chrome on Windows, connected Google Drive app. Affected files are privately owned Google Docs and Drive documents.

Problem: Repeated, inconsistent OpenAI safety-check rejections prevent legitimate, authorized file operations.

The error states:

“This tool call was blocked by OpenAI’s safety checks. Please double check what you are sending.”

Documented examples:

  • October 12, 2026: A larger document backup succeeded while a smaller backup was rejected.

  • October 13: One authorized control-document update succeeded, but the matching update to a second document was rejected.

  • OpenAI Support subsequently confirmed that an October 12 operation passed action approval but was blocked by a separate safety check.

Expected behavior: Authorized file operations should complete reliably, or a technical investigation should determine why they are being rejected.

Actual behavior: Similar operations produce contradictory outcomes, interrupting document preservation and research workflows.

Existing Support case: #17122585

I have provided OpenAI Support with over 50 documented incidents, screenshots, timestamps and error messages. Despite repeated requests, I have not received confirmation of an accepted engineering investigation.

Request to OpenAI staff:

Could someone help obtain a formal technical-escalation decision and have the inconsistent behavior reviewed by an appropriate specialist?

I am not asking for safety protections to be bypassed. I am asking whether these specific rejections are correctly classified and how legitimate connected-app operations can be completed reliably.

All supporting evidence is available privately through the existing Support case.

2

Hey 3ipka3651, welcome to the forum!

That's indeed annoying.

I took a look at the case you shared. As you mentioned there, you haven't tested this in Chrome's Incognito mode yet. Could you give that a try and let us know if the issue persists?

As noted in your case, this is also being treated as product feedback so the engineering team can review it.

In the meantime, I'd recommend using the app as a temporary workaround for this use case.

Avinash

3

Hi Avinash,

Thank you for looking into my case.

I appreciate your suggestion about Chrome Incognito mode. However, the central problem is that ChatGPT’s own safety checks are rejecting authorized Google Drive operations, sometimes while comparable operations succeed.

I’m trying to understand whether this behavior has actually been referred to engineering for investigation, rather than simply recorded as product feedback.

Could you clarify whether an engineering issue has been filed or is being actively reviewed?

I’d be happy to provide additional technical details through an appropriate private support channel, but I want to avoid sharing personal family records publicly.

Thank you for your help.