[Bug] Fable 5 safeguard flag fires on every message in normal conversations, not "sometimes"
Environment
- Claude Code version: 2.1.199
- Platform: macOS (Darwin 25.3.0)
- Model: claude-fable-5
What happens
Every single message in ordinary software-engineering conversations triggers:
This model's safeguards flagged this message. This sometimes happens with safe, normal conversations. Switched to Opus 4.8. Send feedback with /feedback or learn more: https://support.claude.com/en/articles/15363606
The operative word in that message is "sometimes". In practice it is always. Not edge cases, not dual-use security content: routine, safe, normal coding conversations. 100% of messages get flagged and the session silently degrades to Opus 4.8, which makes Fable 5 effectively unusable.
Expected
- Safeguards do not flag routine, safe conversations on every message.
- If a fallback does occur, the messaging does not misrepresent its frequency as occasional.
Comment
Frankly, the "sometimes happens with safe, normal conversations" wording suggests the team does not understand how this classifier actually behaves in production. A safeguard that fires on essentially every benign message is not a safeguard with occasional false positives; it is a misconfigured gate, and shipping it with reassuring copy indicates nobody validated it against real usage.
Related
#73753, #73740, #73041, #72916
3 Comments
Can't wait to put my money towards codex 5.6, it can't come soon enough.
I have same problem, this is absolute garbage, I also send many feedbacks, no return??
I must say, Anthropic did OpenAI great favour. Released model that simply dosent work at all xD Fable is great reciting recipes for cake on max effort, but any other task - it just fires safeguard and fallback to Opus.