[BUG] Erroneous, Completely Unrelated Response

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?

Around 09:15 UTC, when working on a repo, developing an ansible playbook, I asked CC to update the playbook to distribute a public certificate to the nodes that the playbook would run on.

After 5 minutes, it came back with a completely unrelated response, covering GDPR and a massive dataset from Huggingface -- this was not a bad prompt, or something that could be resolved with "best-practices", and looked more like session contamination on the Anthropic side

When asked about the issue, CC responded with the following

I genuinely don't know exactly what caused that. What appeared in my response was clearly not from our session — it contained a massive HuggingFace dataset name, GDPR boilerplate, and a
  competitive programming solution about a grid problem. None of that was anything I generated or intended to send.

  What I can say honestly:
  - I don't have visibility into other users' sessions, and I shouldn't be able to receive content from them
  - That response wasn't something I composed — the structure and content are completely foreign to our conversation
  - It looks like some kind of system-level content injection or caching error, not something I did intentionally

  What you should do: Report this as a bug at https://github.com/anthropics/claude-code/issues. Include:
  - The approximate time it happened
  - That a large block of unrelated content appeared in response to a normal message
  - That it consumed significant context/usage

  The usage impact is a legitimate grievance — you asked a simple question and got charged for a massive erroneous response. That's worth flagging to Anthropic directly.

  I'm sorry that happened. Want to carry on with the SSL cert question now?

This chewed through 26% of my paid usage limits, and while that is annoying, I'm far more concerned about potential session leakage.

Also to note: I have never had a conversation anywhere (on with CC or claude.ai) with anyhting to do with GDPR.

What Should Happen?

The response should have been relevant to the current conversation, and not like a completely different and untreated conversation

Error Messages/Logs

Steps to Reproduce

I have no idea how to reproduce this.

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.97 (Claude Code)

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

WSL (Windows Subsystem for Linux)

Additional Information

_No response_