[Bug] Claude Fable 5 safeguard false positive on authorized security remediation content

Bug Description
Subject: Fable 5 safeguard false positive — authorized security remediation task

Request ID: req_011CdjVAVen2CDVkDsbTRozZ
Model: Claude Fable 5 (via Claude Code)

My message was blocked by the safeguards (AUP link) while I was working on an internal remediation document. This is a false positive: the content is the
technical analysis and fix for vulnerabilities found in an authorized, internally-commissioned VA/PT on my own enterprise platform — defensive work, not
offensive.

The flag was likely triggered by dual-use wording describing the risks being fixed (e.g. "unsalted MD5 hash", "account takeover"). These describe what's
being remediated, not how to exploit it. The proposed changes are purely defensive (output sanitization, password rotation, TLS/CSP hardening).

Please review the safeguard sensitivity for authorized security-testing/remediation work, which is a permitted use — the block interrupts a legitimate
defensive workflow.

Environment Info

• Platform: linux
• Terminal: gnome-terminal
• Version: 2.1.222
• Feedback ID: 79d26922-0f53-4ea3-a28f-68bfd5e1723f

Errors

[]