The Northflank breach means you should treat every secret stored on the platform as stolen, starting with your AI provider keys. Northflank disclosed the incident on October 30, 2026, and told customers to assume its stored environment variables were decrypted. Some users have already reported strangers using their AI keys.
If you host anything on Northflank, rotating your secrets today is the safest move. If you don’t, it’s still a good prompt to check where your own keys live.
Quick facts
- Disclosed: October 30, 2026, after a threat alert from security firm Wiz.
- Northflank’s advice: “assume all environment variables have been decrypted” and rotate all keys.
- Reported impact: unauthorized use of customers’ AI provider keys.
- Still unknown: how attackers got in and how many customers were affected.
That list is short because the company hasn’t shared much more yet. Everything below is what we can say with confidence, plus what to do about it.
What we know about the Northflank breach so far
According to Qovery’s incident write-up, Wiz sent Northflank a threat alert at 11:43 UTC on October 30. Northflank’s status page then told users to rotate environment variables immediately, and at 13:58 UTC it escalated the warning to assume all of them had been decrypted.
Northflank has not published a root cause or a list of affected customers. Resultsense, summarizing a teiss report, says its questions to the company went unanswered. One caveat: Qovery sells a competing platform, so read its framing with that in mind. The timeline itself matches Northflank’s public status updates.
On October 31, Northflank turned on audit logs for every customer, including free accounts. It also said the AI key abuse “may be consistent with the broader pattern of attacks targeting AI credentials.”
Why attackers want your AI keys: LLMjacking
Stolen AI keys let someone run models on your account while you pay the bill. Wiz calls this LLMjacking, and the attacker either burns the credits or resells the access.
The numbers can get ugly. In two cases Wiz investigated, compromised enterprise credentials drove roughly 2.5 million Amazon Bedrock calls and about $100,000 in usage. Your bill won’t look like that, but a single stolen key on a busy plan can still cost real money overnight.
Northflank isn’t alone. Wiz’s “Black October for AI Developer Platforms” report counts at least nine incidents at AI developer platforms last month. We’ve covered two of them: the Railway breach and the OpenCode exposed keys.
What to rotate first after the Northflank breach
Not every secret carries the same risk, so work top-down instead of rotating at random. This order puts the quickest-to-abuse keys first.
| Priority | What to rotate | Why it comes here |
|---|---|---|
| 1 | AI provider keys (OpenAI, Anthropic, Google, and similar) | Users already reported misuse of these, and they can run up a bill within hours. |
| 2 | Payment and billing keys | A stolen payment key can move real money or expose customer data. |
| 3 | Cloud credentials and database passwords | These open the door to your actual infrastructure and data. |
| 4 | Session and JWT signing secrets | Attackers can forge logins if these leak. |
| 5 | OAuth, webhook, Git and registry tokens | Lower odds of immediate damage, but they let someone tamper with your code and deploys. |
Rotating can feel like a lot of work. The payoff is that a rotated key is worthless to whoever stole it, no matter what else they learned.
How to rotate without breaking your app
The common mistake is revoking the old key before the new one is live, which takes your app down. Follow this sequence instead.
- Create the new key at the provider.
- Add it to your app’s environment variables and redeploy.
- Confirm the app works with the new key.
- Revoke the old key at the provider.
Revoking at the provider is the step people skip. Deleting the variable on Northflank doesn’t stop the stolen copy from working.
While you’re there, create the new AI keys with limits. Restrict them to the models you actually use and set a monthly spending cap, so a future theft has a ceiling.
How to check if someone already used your keys
Open each AI provider’s usage and billing page and look at the last several days. You’re hunting for spikes, unfamiliar models, or traffic at hours when your app is quiet.
Then check Northflank’s new audit logs for API keys, users, SSH keys, OAuth apps or webhooks you don’t recognize. If you find any, remove them and rotate again.
Frequently Asked Questions
Was my data stolen in the Northflank breach?
Northflank hasn’t said how many customers were affected, so you can’t rule it out. The safe assumption is that secrets stored on the platform were exposed.
I use Northflank’s bring-your-own-cloud option. Am I affected?
That isn’t clear yet. Ask Northflank directly, and rotate your secrets while you wait for an answer.
What is LLMjacking?
It’s when attackers use stolen AI credentials to run models on your account. They either drain your credits or resell the access to others.
Should I stop using cloud platforms that store my secrets?
Not necessarily, but limit the damage. Use separate keys per project, set spending caps, and keep a list of every platform that holds a key.
The bottom line
Don’t wait for Northflank to explain how it happened. If your keys were on the platform, rotate the AI ones in the next hour, then work down the list this week, and set spending caps on everything new. It’s tedious, but it’s far cheaper than a surprise bill, and the next breach will be easier to shrug off once your keys are scoped and capped.
Leave a Reply