[Bug] Fable 5 safety classifier false positive on benign automation terms
Bug Description
False positive on Fable 5 safety classifier.
My message was flagged for "cybersecurity or biology topics" and the session
was switched to Opus 4.8, but the content was ordinary business automation
work
(marketing content pipeline, email outreach tooling, CRM integration — no
security research or exploit-related content).
Likely trigger: benign technical terms like "API key", "IMAP", "bot",
"session id" used in a normal automation context.
Language: I primarily work in Turkish — if the classifier handles Turkish
text differently, that may contribute.
Expected: message should not be flagged; Fable 5 should stay active.
Environment Info
- Platform: darwin
- Terminal: Apple_Terminal
- Version: 2.1.185
- Feedback ID: e6b31bf5-10dc-4f28-a118-2405e0e8bcd2
Errors
[]This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗