2 related articles

Encountering false positives in Claude Code? Learn how to use the /feedback command, thumbs buttons, and other channels to appeal misclassifications and improve AI safety classifiers.
When AI Flags an Excel Task as a Secur…
A user creating an Excel financial spreadsheet was flagged as a cybersecurity threat by an AI system. This deep-dive examines why automated moderation fails, how appeals systems can be broken, and what it means for trust in AI services.