Bug Report: Repeated False AUP Violations in Claude Code

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Bug Report: Repeated False AUP Violations in Claude Code

Summary

Claude Code repeatedly triggers "Usage Policy violation" errors on completely benign technical messages during a normal software development session.

Environment

• Claude Code CLI (latest)
• Model: claude-opus-4-5-20250514
• OS: Windows 11 / WSL2
• Date: 2026-01-03

Reproduction

During a routine debugging session working on a Telegram bot, the following types of messages triggered false AUP violations:

1. Simple follow-up instructions like "Try /browse now"
2. Technical explanations about API response parsing
3. Numbered lists describing code issues
4. Messages containing code-related terms like "slug", "parsing", "endpoint"

Example Triggers

These innocent messages caused the error:

• "Try /browse now!" (telling user to test a Telegram bot command)
• "It's calling counterpartyService.getCollectionAssetsPaginated()." (describing a function call)
• Numbered lists explaining technical issues with API responses
• Messages containing terms like "data.success" or "rare-pepe vs rare-pepes"

Error Message

API Error: Claude Code is unable to respond to this request, which appears to violate our Usage Policy (https://www.anthropic.com/legal/aup). Please double press esc to edit your last message or start a new session for Claude Code to assist with a different task. If you are seeing this refusal repeatedly, try running /model claude-sonnet-4-20250514 to switch models.

Frequency

Occurred 6+ times in a single 1-hour session, disrupting workflow significantly.

Impact

• Breaks conversational flow during debugging
• Loses context when user has to restart
• Frustrating user experience
• No clear pattern to avoid triggering it

Expected Behavior

Technical software development conversations should not trigger AUP violations. The safety filter appears overly aggressive on benign content.

Suggested Fix

Review and tune the safety filter to reduce false positives on:

• Technical programming terminology
• Bot/CLI command references (e.g., "/browse", "/orderbook")
• API and database discussion
• Code snippets and function names

What Should Happen?

Claude should work normally without a AUP warning on every normal prompt:

API Error: Claude Code is unable to respond to this request, which appears to violate our Usage Policy (https://www.anthropic.com/legal/aup). Please double press esc to edit your last message or start a new session for Claude Code to assist with a different task. If you are seeing this refusal repeatedly, try running /model claude-sonnet-4-20250514 to switch models.

Error Messages/Logs

API Error: Claude Code is unable to respond to this request, which appears to violate our Usage Policy (https://www.anthropic.com/legal/aup). Please double press esc to edit your last message or start a new session for Claude Code to assist with a different task. If you are seeing this refusal repeatedly, try running /model claude-sonnet-4-20250514 to switch models.

Steps to Reproduce

I'm getting the UAP warnings on every prompt now, regardless of the prompts. This is on a session that's been running for over a week with no problems until today.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.0.76 (Claude Code)

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

WSL (Windows Subsystem for Linux)

Additional Information

_No response_