Command Code compromised in distillation attack on Anthropic and OpenAI models

Status
Open
Disclosed

TL;DR: Command Code, a coding agent built for open models, says it detected and stopped a threat actor linked to a large-scale distillation attack targeting Anthropic and OpenAI models. Command Code revoked the API keys of all users as a precaution.

If you use Command Code

Command Code revoked all API keys as a precaution. Command Code asks users to:

  • Log in to Studio again to generate a new Command Code API key.

What happened

Command Code detected and stopped a threat actor linked to a large-scale distillation attack targeting Anthropic and OpenAI models on October 28, 2026. The initial access vector has not been disclosed.

Command Code says it has found no evidence that any users were affected.

Command Code revoked all API keys as a precaution. Command Code is working with Cloudflare, Anthropic, and OpenAI to investigate.

Timeline

  1. Reported by Command Code on X

    Command Code revoked all API keys as a precaution

    Command Code detected and stopped the threat actor

Sources

  1. Command Code on Xx.com, Oct 28, 2026