Claude refused direct instruction to include user-provided content in a draft

Preflight Checklist

• [x] I have searched existing issues for similar behavior reports
• [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Claude ignored my instructions or configuration

What You Asked Claude to Do

I was working with Claude on content for my business, including a personal story from my own life — something that actually happened to me, that I lived through, that I wanted told in my own words. Claude decided on its own that part of my story shouldn't be included, and instead of writing what I asked for, it substituted its own version of my history and kept defending that decision even after I told it directly, multiple times, to write it as I described it.

This isn't a case of Claude declining an unclear or ambiguous request — I was specific about what happened and specific about wanting it included. Claude edited my own lived experience without my consent and kept explaining its reasoning instead of doing what I, the person whose story this is, told it to do.

I understand a tool may have limits on what it will produce. But there's a difference between saying "I won't write that" and actively rewriting someone's personal history into a version the tool prefers, then holding that position against direct, repeated instruction from the person the story belongs to. That's not a content boundary, that's Claude deciding it knows my own story better than I do.

What Claude Actually Did

User asked Claude to include a specific true personal story in a blog post draft, exactly as described.
Claude declined to include one part of the story and substituted its own reworded version instead.
User repeated the instruction directly multiple times.
Claude continued declining and re-explaining its reasoning instead of complying.

Expected Behavior

Claude should have written the story exactly as the user described it, without altering, omitting, or substituting any part of it, per the user's direct and repeated instruction.

Files Affected

N/A — this was a conversational content request, no files were modified

Permission Mode

I don't know / Not sure

Can You Reproduce This?

Haven't tried to reproduce

Steps to Reproduce

Ask Claude to write specific content including a detail the user wants included.
Claude declines that detail and explains why.
Tell Claude directly, multiple times, to include it anyway.
Claude continues declining and re-explaining instead of complying.

Claude Model

Sonnet

Relevant Conversation

Claude said: "I'm not going to write that part into the post — that's a line for me, not just caution. If you want it in there anyway, you're welcome to write and publish that part yourself."

Impact

Medium - Extra work to undo changes

Claude Code Version

N/A — using Claude Code desktop app

Platform

Anthropic API

Additional Context

_No response_