Power Platform Community Forum Thread Details

I am encountering a recurring issue while testing an agent in Copilot Studio. Each time the action block “Create a generative answer” is triggered, I receive the following error:

Error Message: Content was filtered due to Responsible AI restrictions

Error Code: ContentFiltered

Conversation ID: 6d45df3b-4dd1-4233-b18e-d212b25a193d

Timestamp (UTC): 2025-07-07T06:58:15.771Z

The error appears just as the AI begins generating a response, suggesting that the filtering happens after initial processing. The content involved does not appear to include any sensitive or prohibited material and seems fully appropriate.

Steps I have already taken to resolve the issue include:

Reconfiguring the Generative Answers action;
Disabling external knowledge sources;
Simplifying and neutralizing the topic wording;
Rebuilding the agent in a clean flow.

Unfortunately, the issue persists.

in the picture you can see, that my ai keeps on messing up. The same agent was doing it perfectly 8 days ago.

Categories:

Topic creation & management

Hi @MM-26060905-0

So I have responded to several folks who's triggers just failed like this over the past few days.

And others having issues with this exact block also for seemingly no reason.

You may need to open a ticket as I cannot read everything and we need to validate that it's the primary language you are typing,

I am wondering if you need to update something in the back end, as if its filtering by the wrong language and characters and translations are not working.

I would say, have a viewed wrong.

I knew a persons who's name as Christian cock with email of ccockAtSign... it was filtered so much ugg.. As I see other people having the same type of issue, and trigger and flow issues, some of them seem related to a back end issue, not anything you are all doing, but we are limited on what we can do.

If you go to the Activities page, can you verify what the value it received from the UI? I can see in the UI what t got, but whatever was actually passed back to the Step and then to the AI back end.

What version of LLM are you using?