Northflank breach: assume all secrets decrypted, AI keys abused
Northflank told customers on Oct 30 to assume every environment variable was decrypted and rotate all secrets after AI API keys were used without authorization. No root cause yet.
Cloud platform Northflank told every customer to assume all of their environment variables have been decrypted and to rotate every secret immediately, after an intrusion that led to unauthorized use of customers' AI API keys. The disclosure runs on Northflank's incident status page, updated through October 30 and into October 31. Root cause, scope, and whether data was exfiltrated are still unstated.
If you deploy on Northflank, this is a today task, not a this-week task.
Timeline
The status page logs the response in UTC:
- 11:43 — Northflank receives a threat alert from Wiz; engineers review it one minute later.
- 12:02–13:05 — A network-policy patch, then two further patches, are deployed; Northflank requests Wiz CIRT support.
- 13:40 — First customer notice: "Rotate environment variables immediately."
- 13:58 — Escalation: "Assume all environment variables have been decrypted. Disable/rotate all keys."
- 15:12 — Secret rotation for Postgres and MongoDB enabled for all users. The page adds: "Multiple users are confirming that their AI keys have been used."
- 23:38 — "Users have reported unauthorised use of AI API keys supplied to the Northflank platform."
- Oct 31, 01:04 — Audit logs enabled for all customers, including free users.
- Oct 31, 13:34 — MySQL and Redis rotation made available to all customers.
Environment variables on a platform like this routinely hold database passwords, cloud credentials, auth tokens, and AI provider keys. Northflank's guidance is to treat all of them as exposed.
What's confirmed, what isn't
Confirmed: unauthorized use of customers' AI API keys, reported by multiple users. Northflank has not said how the environment variables were decrypted, how many customers are affected, how long exposure lasted, or whether secrets were exfiltrated in bulk. The investigation is described as ongoing with "no further confirmed findings." The status page points to Wiz's "Black October" report for overlapping attacker IP addresses — attribution stops there; no group is named.
Action checklist
- Rotate every secret stored on Northflank — environment variables, secret groups, global secrets, and anything referenced in workloads. Northflank's own PostgreSQL, MongoDB, MySQL, and Redis rotation flows are now live for all accounts.
- Revoke and reissue AI provider keys (OpenAI, Anthropic, Google, Bedrock, etc.) at the provider, not just in Northflank, then update the workloads that use them.
- Pull provider usage and billing logs and look for unexpected inference spend — the observed abuse pattern across these incidents is large volumes of API calls run on stolen keys.
- Disable public access to add-ons where you enabled it.
- Turn on and review the audit logs Northflank just enabled, and report any suspicious activity to Northflank support and your AI providers.
Context
This is not an isolated event. It's the latest in a cluster of October intrusions at platforms that hold customer secrets for AI workloads — what Wiz has labeled "Black October for AI developer platforms." We've already covered Railway, OpenCode's Console, and Warp from the same window; a Cloud Security Alliance research note tracks Zeabur (intrusion October 14–15) and Fireworks AI in the same pattern. The common thread is a single-tenant secret store holding high-value, high-liquidity credentials — AI keys that convert directly to compute spend — behind a platform whose blast radius is every customer at once. Wiz has not confirmed a single actor behind all of them. The defensive takeaway is the same each time: secrets handed to a third-party deploy platform are only as safe as that platform's weakest day, and rotation has to be a drill you can run in minutes.