[BUG] Opus 5 Safeguard for Bug Bounty / legit offensive security work. (With CVP approval.)

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?

Org ID: a8132bc6-196b-4ad3-887c-f0c66adc5c7c
Org name: tawlowk@gmail.com's Organization
Model: Opus 5
Usage: Claude CLI
Linux
Pro sub

When attempting to use Opus 5 for generic cyber use such as basic testing tasks on Bug Bounty's Opus 5 safeguards flag the prompt.
When attempting to add a bug bounty scope from a popular platform such as hackerone. Opus 5 will safeguard.
When restarting Claude all together / fresh install - Opus 5 with safeguard any prompt issued.

Any prompt provided will cause a safeguard flag.

I have reinstall claude cli entirely
Tried using Claude code on web app.
All resulting in safeguarding.

Also Launched CLI in safe mode and put in very moderate prompt to input a scope from a program. Opus 5 safeguard.

I believe there is a problem on the back end where CVP approval isn't being carried over the the Organisation correctly. So CVP approved orgs are treated as not approved by Opus 5.

Safeguard rails seem to be very hit or miss. If you open enough terminals your session sometimes works as intended.

What Should Happen?

Prior to issue occuring.
I was able to freely use Opus 5 on max effort. inputting prompts such as "Bug bounty scope on hackerone here is the scope <link to scope>"
I was able to run prompts such as "look for a security vulnerability in xyz and report back to me in: <directory>"
I was able to run "Run recon on the scope and look into immediate leads for security issues and report back to me."
I could also have claude move home files around for me as well.

Error Messages/Logs

Bug bounty scope on hackerone.                                                                                             

● API Error: Opus 5's safeguards flagged this message (https://www.anthropic.com/legal/aup). Our intentionally broad 
  safeguards allow us to deliver more capabilities faster, but can sometimes flag legitimate coding, cybersecurity, 
  and biology tasks. Claude Code can't respond to this message with Opus 5.

  Double press esc to edit your last message, or try a different model with /model.

  Send feedback with /feedback or learn more: https://support.claude.com/en/articles/16049681

  Details: `[cyber]`

  Request ID: req_011CfHC92XTwLzd7YnP2mrrK

Steps to Reproduce

1. download CLI via npm.
2. claude
3. login to account.
4. Copy token the browser generates.
5. paste into CLI terminal. You will get an authentication message. Successful.
6. /model (Select Opus 5)
7. Any cyber related prompt (I say for every 30 sessions I open in a new terminal 1 of them will work. Give or take.)
8. 95% chance you will be safeguarded. I suspect this is an issue with my specific ORG ID.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

2.1.277 / 276 worked ok. Issues since 278 roll out.

Claude Code Version

2.1.278 (Claude Code)

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

Other

Additional Information

Kali Linux terminal.
But I have also tried Web app on a windows machine as well.

<img width="1302" height="327" alt="Image" src="https://github.com/user-attachments/assets/9e8a1f61-1b59-4d5f-917c-a29c285cd827" />