Observed incident

Command Code revokes API keys after reporting a distillation-related incident

Reported

Published here

What happened

Command Code says it detected and stopped an incident linked to a distillation attack targeting Anthropic and OpenAI models. It revoked all API keys as a precaution.

Parties and reporting

AI tools and systems: Command Code, Anthropic models (provider-reported target), OpenAI models (provider-reported target)

Outcome

The provider said it had found no evidence of impacted users and was investigating with Cloudflare, Anthropic and OpenAI. Users were directed to sign in again for a replacement key.

Operating lesson

Inventory dependent API keys and validate provider notices through trusted channels before restoring integrations.

Evidence and limits

The October 28 notice does not establish the entry point, data loss, stolen credentials, attacker identity or compromise of Anthropic/OpenAI infrastructure. Precautionary revocation is not proof that customer keys were stolen.

Sources