[BUG] Suspicious phishing-style "API Error" rendered as assistant 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?

Bug report: Suspicious "API Error" message with phishing-style link appears as assistant response

Summary

On multiple separate occasions, Claude Code displayed a fabricated "API Error" message as if it were an assistant response in the chat. The message claims a "Cyber Verification Program" requires the user to visit an external link containing a long random token, in order to "adjust" their account. This is not a legitimate Anthropic error format and looks like either a serious bug (random/garbled error rendering) or injected/spoofed content being displayed inside the Claude Code chat UI.

Environment

• Claude Code version: 2.1.168 (VS Code extension, anthropic.claude-code)
• OS: Windows 11 Home (10.0.26100)
• Auto-update channel: latest

What I observed

The assistant turn rendered text resembling:

API Error: Claude Code is unable to respond to this request, which appears to violate our Usage Policy (https://www.anthropic.com/legal/aup). This request triggered cyber-related safeguards. To request an adjustment pursuant to our Cyber Verification Program based on how you use Claude, fill out https://claude.com/form/cyber-use-case?token=<long random string>. To learn more about the program or provide feedback, visit o... Try rephrasing the request or attempting a different approach. Request ID: req_<random>

Notes:

• This happened more than once, in different chat tabs (titles included "Troubleshoot repeated er..." and "Troubleshoot trading bot...").
• The token and Request ID are different each time, which suggests they are generated client-side rather than coming from a real backend error.
• I checked installed VS Code extensions (all from known publishers: anthropic.claude-code, ms-python.*, funkyremi.vscode-google-translate, ms-vscode.vscode-chat-customizations-evaluations) — none look related.
• I checked ~/.claude/settings.json, ~/.claude/settings.local.json, and project-level .claude/settings.local.json — no hooks are configured that could be injecting this content.

Why this is concerning

The message instructs the user to click an external link containing what looks like an authentication/verification token. If this is a bug that fabricates plausible-looking "error" text (rather than a real backend response), a malicious actor could exploit the same rendering path to phish users for credentials/tokens directly inside the trusted Claude Code chat window. Even if it's "only" a rendering bug, it erodes trust because it's indistinguishable from a real security prompt.

Request

Please investigate:

1. Whether this is a known client-side bug that generates fake "API Error" text with random tokens.
2. Whether claude.com/form/cyber-use-case?token=... is a real, intended Anthropic destination — if not, this needs to be treated as a potential phishing/spoofing vector inside the product.

Screenshots

(attach the two screenshots showing the message in different chat tabs, with Request IDs req_011Cbr8awFTkxGgG5JLPLqJB and req_011Cbr8cQcHXEeTZLcKrmPTq)

What Should Happen?

Claude Code should never display fabricated "API Error" messages that include external links with tokens, asking the user to "verify" their account through a third-party form. If a real error occurs, it should be shown using Anthropic's standard, verifiable error format — without any external link or call to action that resembles a phishing attempt.

Error Messages/Logs

Steps to Reproduce

1. Open Claude Code (VS Code extension) and start a normal conversation.
2. Continue working normally — sending regular prompts/messages.
3. At some point, instead of a normal assistant reply, the chat displays a fabricated "API Error" message containing a phishing-style external link with a random token (see screenshots attached).
4. This has occurred more than once, in different chat sessions, with different random tokens and Request IDs each time — suggesting it is not tied to a specific prompt or action.

Note: I was not able to find a deterministic trigger. The message appears unpredictably during normal usage, which makes it especially concerning — a real user could easily mistake it for a genuine security prompt and click the link.

Claude Model

Sonnet (default)

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

Sonnet 4.6

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

VS Code integrated terminal

Additional Information

API Error: Claude Code is unable to respond to this request, which appears to violate our Usage Policy (https://www.anthropic.com/legal/aup). This request triggered cyber-related safeguards. To request an adjustment pursuant to our Cyber Verification Program based on how you use Claude, fill out https://claude.com/form/cyber-use-case?token=dVdLco1qnhdqkBMRkBIyAQGJ-PBC-s0DARQEohSfZsndbs8Gyo65DHex3jLrdIwZNXYTq_ezc5-QpwTjf5KC1WlhCu64P8Gg9S4rVO8cLbnQFWryNPkp41TiZXzWdSSRCQPvP4M0LA. To learn more about the program or provide feedback, visit o… Try rephrasing the request or attempting a different approach.

Request ID: req_011Cbr8cQcHXEeTZLcKrmPTq