Cloud application hosting platform Northflank is warning customers to rotate all credentials and secrets stored on its platform after a security breach led to unauthorized use of customers' AI API keys. The company disclosed the incident on October 30.
What happened
Northflank disclosed a security incident on October 30, warning customers to immediately rotate their environment variables and API keys after receiving a threat alert from cloud security company Wiz.
Assume all environment variables have been decrypted. Disable/rotate all keys.
Northflank later confirmed that multiple customers had reported unauthorized use of AI API keys stored on the platform. The advice to customers covers all credentials and secrets held on the platform, not only the keys that were reported as abused.
How it was done
The method is not public. The company has not disclosed how attackers gained access to its systems, how many customers were affected, or whether attackers stole or abused credentials other than the reported AI API keys.
What is known is the shape of the exposure: secrets that customers stored on the hosting platform, including environment variables, are being treated as decrypted. The abuse that customers reported was of AI API keys.
What to check
If you run workloads on a hosting platform that stores your environment variables, work through the following.
Inventory every secret held by the provider
Export the environment variables and secret references for each service and environment, including preview and staging deployments that people forget. Diff that list against your own secret manager inventory. Anything present on the platform but absent from your records is an unrotated credential you are not tracking.
Rotate AI provider keys first
In your model provider consoles, issue new keys, update the running services, then delete the old keys rather than leaving them disabled. Confirm deletion by calling the API with the retired key and checking that you receive an authentication error.
Pull usage and billing data for those keys
Open the usage dashboards for each AI provider and compare request volume, model mix and spend against your normal baseline. Unfamiliar models, requests outside your working hours, or traffic from regions you do not deploy in are signs of third party use.
Rotate the rest, not just the obvious ones
Cover database passwords, object storage keys, webhook signing secrets, OAuth client secrets, SMTP credentials and registry tokens. Track each one in a checklist with an owner, since partial rotation leaves a working path back in.
Look for persistence created with stolen keys
Check your cloud and SaaS admin consoles for new API keys, service accounts, OAuth grants and webhook endpoints created recently. Anything you cannot tie to a change request or a named engineer should be removed and investigated.
Add spend and rate limits to model APIs
Set hard monthly caps and per key rate limits in each AI provider console, and wire budget alerts to a channel someone reads. Without a cap, abuse of a leaked key shows up first as an invoice.