The safety classifier blocked subsequent benign queries and redirected him to Claude Opus.
Am I using Fable wrong somehow, or do I just look like a dangerous person who thinks about security too much based on Claude memory?
I asked Fable to check whether a potential logo looked like any other logos, and it punted me back to Opus as a security risk.
Users disapprove of Fable AI flagging even innocuous questions as security risks because the behavior affects basic queries like those on high school biology.
No Digg Deeper questions have been answered for this story yet.
Admittedly my very first query to Fable was to find me some historical CVEs of a particular flavor. Maybe the weaker classifier models are looking at this history too, and I’m suspicious from the start?
Am I using Fable wrong somehow, or do I just look like a dangerous person who thinks about security too much based on Claude memory?
I asked Fable to check whether a potential logo looked like any other logos, and it punted me back to Opus as a security risk.
@geoffreyirving Kind of wouldn’t surprise me. I have extremely innocuous questions that get flagged. Basically anything involving high school biology. If I instruct Fable to search my chat history that usually triggers it too.
@geoffreyirving It probably sees this in your history and punts you for that