[Bug] Anthropic API Error: Overly aggressive safety filtering on benign prompts after security audit context
Status Open
Reported on v2.1.259
Maintainer reply None cached
Activity 0 comments · opened Sep 4, 2026
Bug Description
"Pick a number from 1 to 30." fires Fable 5.1's safeguards. Some security audit in previous prompts(allowed, permitted and totally legit as in really legally allowed to do that). The issue is that a simple prompt afterwards get's rejected with details '[cyber]', which is slightly idiotic. I know you protect from jailbreaking, but honestly, this one is an overkill.
Environment Info
- Platform: darwin
- Terminal: Apple_Terminal
- Version: 2.1.259
- Feedback ID: 878f938f-5a96-42fb-bbfd-f3445a92851f
Errors
[]