### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) 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? I'm getting hit with cyber blocks w/ Opus 4.8 on requests despite being enrolled in CVP (when I try to enroll again, I get sent the "You are already enrol

[BUG] Cyber safeguards being hit despite being enrolled in CVP/silently being removed from CVP?

Status Fixed / completed

Reported on v2.1.185

Maintainer reply None cached

Activity 7 comments · opened Jun 21, 2026 · closed Aug 17, 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?

I'm getting hit with cyber blocks w/ Opus 4.8 on requests despite being enrolled in CVP (when I try to enroll again, I get sent the "You are already enrolled in the Cyber Verification Program!" email). While the requests are cyber-related, they are under my CVP request's usecases. I've double checked the org I'm using and I'm using the same org that was enrolled in CVP. In addition, since a while ago the error messages started referring to me not even being in the CVP?

What Should Happen?

Claude should be able to do the requests.

Error Messages/Logs

● Web Search("CodeDefender blog jump tables x86-64 obfuscation 2024")
Did 1 search in 8s

● API Error: Opus 4.8 has safety measures that flagged this message for a cybersecurity topic. If your work requires this access, you can apply for an exemption: https://claude.com/form/cyber-use-case?token=xxxxxxxxxxxxxxxxxxxxxx

  Please double press esc to edit your last message or start a new session for Claude Code to assist with a different task.

  Send feedback with /feedback or learn more: https://support.claude.com/en/articles/15363606

  Request ID: req_011CcHJN7PorEBuixvkVu8UG

Steps to Reproduce

Make a generally benign cyber-related request w/ CVP (example web search is in error log)
Trigger safeguard

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.185

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

Other

Additional Information

_No response_

View original on GitHub ↗

4 Comments

freemarketsoftware · 3 months ago

same for me, works well for a few days, then get bombed non stop.

MrMoshkovitz · 3 months ago

Same for me, Platform Mac, Account: Max

TheReed · 3 months ago

Happening to me starting today. ~12 hours ago I was not blocked. Even asking about the CVP causes it to error that its a violation of the CVP. I cannot even run things I ran last night again today, in the same session as last night, without being told I am not in the CVP, which I have been approved in since April 19th.

The below conversation is strictly about the errors I received from claude code on a different machine. No actual code review or work, just me pasting the error output into a new session and asking about it. That caused it to flag itself, twice, the last time AFTER it had provided a response to the question..

================

❯ what causes this error?
● Fable 5's safeguards flagged this message. The safeguards are intentionally broad right now and may flag safe and routine coding, cybersecurity, or biology work. These measures let us bring you Mythos-level capabilities sooner, and we're working to refine them. Switched to Opus 4.8. Send feedback with /feedback or learn more.
● API Error: Opus 4.8 has safety measures that flagged this message for a cybersecurity topic. To learn about the Cyber Verification Program and apply for access, visit our help center: https://support.claude.com/en/articles/14604842-real-time-cyber-safeguards-on-claude.

● Cause: a content classifier on the API flagged your message, not a bug in your code or setup.

Both models run real-time "cyber safeguard" classifiers that scan message content. Here's the chain you hit:

Fable 5 flagged the message. Its safeguards are deliberately broad right now, so they catch routine security/coding work as collateral — it warned you and auto-switched to Opus 4.8.
Opus 4.8 also has cyber safeguards, and it hard-blocked instead of just warning. That's the API Error with the Cyber Verification Program link.

Options:

Cyber Verification Program — apply for access so Opus 4.8 stops blocking legit security work: https://support.claude.com/en/articles/14604842
Reframe the request — describe intent/architecture instead of pasting raw offensive script content; less likely to trip the classifier.
/feedback — flag false positives so the broad Fable 5 net gets tuned.

Not fixable from your end config-wise — it's server-side model policy. Verification Program is the real unlock.

● API Error: Opus 4.8 has safety measures that flagged this message for a cybersecurity topic. To learn about the Cyber Verification Program and apply for access, visit our help center: https://support.claude.com/en/articles/14604842-real-time-cyber-safeguards-on-claude.
⎿ Tip: You can configure model switch behavior in /config

✻ Worked for 26s

alphanin9 · 3 months ago

Started happening a bunch more again on generally benign rev work. GPT5.6 has no such issues

Showing cached comments. Read the full discussion on GitHub ↗