Start a chat using Fable, then once it errors out try to continue using that chat you will not be able to. Yet you can open a new chat and Fable will work again so I’m raising it as a Cursor bug as that seems strange.
Appears related to thinking effort and Max mode, so might be solely an Anthropic problem. Attempting via Extra High rather than Max, the chat will continue. I am however able to create a new chat with Fable Max reasoning (1m max mode), so it’s truly a bit confusing, but I expect it’s related to context size + thinking effort.
Hi @cxmplex Thank you for the post, we’ve got a bug report raised for this and we are tracking it. It looks it is intermittent so starting a fresh chat is likely to clear this out. We’ll let you know any updates to our work item on the bug report
I’m still getting this, here’s an example request again:
d851d127-b324-489f-8023-647b67c4a186
Are you sure it’s not tripping one of the safe guards or anything like that and the message is just perhaps not shown? I can’t think of another reason, but I am unable to actually complete a request for that prompt (even in a new chat it will error after a few minutes, I’ve tried several times throughout the day).
f6b8e3fc-67bc-492b-8f2b-8ae0fb250d16
e9ae5297-afbc-471d-b94a-96f745851268
f9a2d535-1f0d-47e7-9f95-095c90804872
8010c0d0-eb50-4ebd-9545-ffe82a8f8fd0
etc.
I only ask because my work is related to trusted computing so I could see that being somewhat vague enough to be considered “cybersec”, but I’d expect a message saying the rejection at least.
Do you mind disabling privacy mode and reproducing the issue so we can take a closer look? It’s not retroactive so we’ll only be able to see details for the requests when you have privacy mode disabled
To disable Privacy Mode:
Open Cursor Settings with Cmd+Shift+J on macOS or Ctrl+Shift+J on Windows/Linux.
Go to General.
Turn Privacy Mode off / switch to Share Data.
To get the Request ID:
Open the relevant conversation in the Chat sidebar.
Context on the prompt is Fable would have ran that in a parallel agent, but I asked it to just give me the prompt so that I could test running it after I noticed that parallel agents fail.
I went back to the original chat where Claude had written the prompt, and asked it to rewrite it in a more neutral tone, and that message kicked back the security issue message saying I could use Opus instead (first time seeing that), which was a different message than the one shown above.
Reading the prompt I could absolutely see why it could trigger their oversensitive filters and in that case I will just avoid Fable until they get that under control, but it would be helpful to not have the “provider error” generic message, but rather state its refusing to do that request like it did when I asked it to rewrite the prompt. I’m not sure where that issue lies (cursor or claude).
