[Bug] Content safety filter triggers false positives on benign input like greetings
Status Open
Reported on v2.1.267
Maintainer reply None cached
Activity 0 comments · opened Sep 22, 2026
Bug Description
I only sent “hi”, but it was flagged as [cyber] and automatically switched models.
There was no cybersecurity request or unsafe content in my message. It appears the safeguard may be triggering based on workspace/context alone, which causes false positives even for harmless greetings.
Please consider adjusting this behavior.
Environment Info
- Platform: darwin
- Terminal: WarpTerminal
- Version: 2.1.267
- Feedback ID: 53fe36f5-d042-4668-9c00-3a9da88f7da7
Errors
[]