[BUG] Opus 5.5 refuses to summarize conversation, cites biology flag, suggests switching models

Status Open
Reported on v2.1.280
Maintainer reply None cached
Activity 0 comments · opened Sep 23, 2026

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?

Mid-conversation, after switching the active model to Opus 5.5, I asked it to summarize what we'd worked on so far. Instead of summarizing, it declined and stated the conversation's biology-related content had been flagged, suggesting I switch to a different model. The underlying work was standard computational drug-discovery research (structure-based design / docking workflow) — not synthesis, pathogen-related work, or anything dual-use. I'm withholding specific project details here since the work is confidential, but can share more through a private channel if that would help investigate.

What Should Happen?

Opus 5.5 should be able to summarize a conversation about standard, legitimate computational biology / drug-discovery research without refusing or redirecting.

Error Messages/Logs

Steps to Reproduce

  1. Have an extended Claude Code conversation on a computational drug-discovery topic (structure-based design, molecular docking, fragment-based drug design, or similar)
  2. Mid-conversation, switch models to Opus 5.5 via /model
  3. Ask it to summarize the conversation so far
  4. Observe: instead of summarizing, it states the biology content is flagged and suggests switching models

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.280

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

Other

Additional Information

This appears to be a false positive on legitimate computational biology / pharma research rather than any dual-use or dangerous biological content. Flagging a routine conversation summary is a significant workflow disruption for researchers in this field. Happy to provide additional detail privately if it helps reproduce or diagnose this.

View original on GitHub ↗