Explainer / Published 2026-10-31

Northflank breach let attackers use customers' AI API keys

Northflank told customers to assume every environment variable had been decrypted and to rotate all secrets, after reports of unauthorised use of AI API keys.

Flat editorial illustration of eight coloured key plates mounted in a row on a plain wall, seven with keys hanging from them and one empty.
Conceptual editorial illustration of key hooks and a missing key; it does not depict any company or person.

THE SHORT VERSION

  • Northflank warned customers on October 30 to rotate all environment variables and API keys after a security breach, following a threat alert from the cloud security firm Wiz.
  • The company told users to assume their environment variables had been decrypted and their secrets stolen, and confirmed multiple reports of unauthorised use of AI API keys.
  • Its guidance includes rotating user-provided secrets with the issuing providers, updating workloads and disabling public access to add-ons.

What Northflank told customers

Northflank, a cloud platform for deploying applications, containers, databases and AI applications, warned customers on October 30 to rotate all credentials and secrets stored on its platform after a security breach led to unauthorised use of customers' AI API keys. BleepingComputer reported the warning and the sequence that led to it.

The trigger was a threat alert from the cloud security company Wiz. After receiving it, Northflank told all customers to assume their environment variables had been decrypted and that the secrets stored within them were stolen, with a blunt instruction: assume all environment variables have been decrypted, and disable or rotate all keys.

The guidance is broader than a password reset

The company's status page sets out what customers should do and leaves little room for interpretation. Users who have not already rotated their user-provided secrets should do so immediately, revoking existing credentials with their issuing providers, generating replacements and updating the affected secrets and workloads.

The guidance goes further than the keys a customer typed in. It asks users to assume all environment variables have been compromised and to rotate all secrets and credentials, including those in workloads, secret groups and global secrets, and recommends disabling public access to add-ons. It also notes that secret rotation has been enabled on Postgres and MongoDB for all users.

Why AI API keys are the target

Environment variables are where applications keep the credentials they need to run: database passwords, cloud access keys, authentication tokens and API keys. A platform that hosts many applications holds a concentration of them, and a single compromise can expose credentials for services the customer never connected to the platform's own systems.

AI API keys are especially attractive because they are metered spending. A stolen model key can be used to run inference at the owner's expense, which is a direct financial loss rather than only a data risk, and the usage can be hard to distinguish from legitimate traffic until the bill arrives.

The wider pattern the notice points to

Northflank's status page mentions that the reports may be consistent with a broader pattern of attacks targeting AI credentials, linking to an article on the subject. That framing matters: it suggests the breach is being read as part of a campaign against developers' AI infrastructure rather than as an isolated incident.

Whether that reading holds will depend on the investigation, which the company says is ongoing with no further confirmed findings to share. The notice is careful to distinguish what customers have reported from what the company has confirmed, and that distinction is worth preserving when the story is repeated.

What customers are left to do

For a customer, the practical work is a rotation exercise across every service that ever had a credential stored on the platform. That includes keys used by background jobs, build pipelines and integrations that may not appear in the main application configuration, which is why the instruction to assume all environment variables were decrypted is the important one.

Rotating a key is only half the task; revoking the old one at the provider is the half that stops the abuse. A customer who replaces a key without revoking the old one has changed nothing, and the notice's wording, revoke existing credentials with their issuing providers, is precise about the order.

What a platform breach means for tenants

A hosting platform is a multi-tenant system, and a breach there is not confined to the provider. The credentials a customer stores belong to services elsewhere: cloud accounts, payment processors, model providers. The platform is the place they were kept, not the place that issued them, which is why the notice insists on revocation at the issuing provider.

That distinction also shapes the blast radius. A customer who used one key for several workloads may have exposed all of them at once, and the attackers did not need to compromise those other services directly. Concentration is the business model of a platform and the security problem of one.

Why the notice reads like a checklist

The language of the notice is deliberately absolute. Telling customers to assume decryption rather than to wait for confirmation is standard practice in incident response, because the cost of rotating a key that was not stolen is small compared with the cost of leaving a stolen one live.

It also protects the company legally and reputationally: a notice that customers can act on is better than one that is precise but late. The open question is what happens after the rotation, when the investigation produces findings that either confirm the worst case or narrow it.

What is established, and what is not

Established: Northflank warned customers on October 30 to rotate all credentials and secrets; the warning followed a threat alert from Wiz; the company said to assume environment variables were decrypted; multiple customers reported unauthorised use of AI API keys; rotation was enabled for Postgres and MongoDB. Not established: how the attackers got in, how many customers are affected, and whether any customer data beyond credentials was accessed.

For readers, the useful signal is that the credentials behind AI features are now a target in their own right. This site has covered a sandbox escape in a shared ChatGPT link, a trojanised coding extension and the verification tiers proposed for high-risk models; all three describe the same shift, where the access token is the thing worth stealing.

Sources & reporting limits

  1. Northflank security breach leads to abuse of customers' AI API keys
  2. Northflank status: incident guidance
Keep the thread.

Follow Industry to find related stories on your next visit.

THE WEEKLY EDIT

The week, made clear.

Stories worth understanding, brought together in one weekly edition.

Share this page

Instagram, TikTok & other apps

Copy the page link, then paste it where your chosen app supports links. You choose whether to send or publish.