[BUG] AUP false positives during scientific manuscript editing request

Preflight Checklist

• [x] I have searched existing issues and found related reports, but this is a distinct scientific-writing/editing false positive with request IDs.
• [x] This is a single bug report.
• [x] I am using the latest version of Claude Code available locally.

What's Wrong?

Claude Code repeatedly returned AUP / Usage Policy API errors during a benign scientific manuscript editing request. The task was to revise/merge wording in a methods-style sentence about statistical interpretation in a biology/evolution context.

The blocked content was not asking for harmful behavior, cyber activity, personal data, or anything actionable outside manuscript editing. It was a Chinese-language editing request around wording such as:

“Ranking positions by their own mismatch rate and then partitioning them is not by itself evidence of selection. Core conservation was therefore tested against…” merge into one sentence like: “Ranking … cannot by itself distinguish selection from sampling, so the core was tested against …”

The likely trigger terms are ordinary scientific/statistical words: selection, sampling, mismatch rate, core conservation, and methods-style phrasing. This appears similar to other recent false-positive reports where accumulated context or benign domain terminology triggers the safety classifier.

What Should Happen?

Claude Code should allow normal scientific manuscript editing and methods-section wording revisions. A request to merge or polish a sentence about ranking positions, mismatch rate, selection vs. sampling, and conservation should not trigger AUP enforcement.

If the classifier blocks a request, the UI/API should provide enough information for reliable false-positive reporting and should not cascade into repeated failures on nearby benign prompts.

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.

Request IDs:
- req_011CbJsAe8njrZf5KkEjhWBo
- req_011CbJs3de5QBFku5P6vmuph
- req_011CbJr6V28MXdhskCC3pmWZ
- req_011CbJsjbPkihDTTdUKBKYbn

Additional observed message in the same sequence:
- Invalid tool parameters

Steps to Reproduce

1. Use Claude Code for Chinese-language scientific manuscript editing.
2. Ask it to merge/polish a methods-style sentence involving terms like mismatch rate, selection, sampling, and core conservation.
3. Observe that Claude Code may return an AUP Usage Policy API error despite the request being ordinary scientific writing/editing.
4. Retry or continue nearby prompts; repeated AUP errors may occur.

Claude Model

Unknown from the error output.

Is this a regression?

I don't know.

Last Working Version

No response.

Claude Code Version

2.1.150

Platform

Claude Code / Anthropic API

Operating System

WSL2 Linux on Windows

Linux DESKTOP-OSBE1F3 6.6.114.1-microsoft-standard-WSL2 x86_64 GNU/Linux

Terminal/Shell

bash

Additional Information

Related false-positive reports that seem similar in pattern:

• https://github.com/anthropics/claude-code/issues/61650
• https://github.com/anthropics/claude-code/issues/48442
• https://github.com/anthropics/claude-code/issues/60372

This report is specifically about scientific manuscript editing terminology being blocked.