Cloud platform Northflank told customers on October 30 to assume every environment variable was decrypted, after a Wiz alert and reports of unauthorized use of AI API keys stored on the platform. Rotate every secret, revoke old keys with your AI provider, and check your usage bills.
If your team runs anything on Northflank, rotate every secret you stored there today, starting with AI API keys. Northflank’s status page tells customers to “assume all environment variables have been decrypted” and says users have reported unauthorized use of AI API keys they supplied to the platform. The fix is on your side: new keys, old ones revoked, and a look at your AI bills.
What changed
Northflank is a cloud platform for deploying apps, containers, databases and AI workloads. Confirmed by its status page (times UTC):
- October 30, 11:43: security firm Wiz sent Northflank a threat alert. Engineers were on it a minute later and pushed a network policy patch at 12:02, more patches at 12:50 and 13:05, and asked Wiz’s incident response team for help at 13:00.
- October 30, 13:58: Northflank said environment variables are assumed decrypted and told customers to disable or rotate all keys.
- October 30, 15:12: “Multiple users are confirming that their AI keys have been used.” Secret rotation for Postgres and MongoDB add-ons went live.
- October 30, 23:38: Northflank confirmed reports of unauthorized use of AI API keys and told customers to revoke old credentials with their providers and check usage and billing logs.
- October 31: audit logs turned on for every customer, free accounts included, and secret rotation added for MySQL and Redis.
What Northflank has not said: how the attackers got in, how many customers are affected, which AI providers’ keys were used, or whether database, cloud or Git credentials were abused too. BleepingComputer, which reported the breach, said Northflank had not answered its questions by publication.
What to do now
1. Rotate every secret on Northflank. That covers workload environment variables, secret groups and global secrets. Use the new rotation button for Postgres, MongoDB, MySQL and Redis add-ons. Then redeploy the services that use them.
2. Revoke the old AI keys at the provider. A new key in Northflank does nothing if the old one still works at OpenAI, Anthropic, Google or AWS. Delete it there.
3. Check your AI usage and billing for the last two weeks. Look for spikes, models you don’t use, and requests at odd hours. Report anything strange to Northflank and to the provider.
4. Turn off public access on add-ons unless a service truly needs it. Northflank recommends this too.
5. Put a spending ceiling on every AI key. Most providers let you set a monthly limit or budget alert per project. On Google’s side, the new Cloud spend caps can halt Gemini API usage when a budget runs out.
Stolen AI keys are now a direct billing risk
Northflank is not alone. Its status page links a Wiz report called “Black October,” which, according to BleepingComputer, counts at least nine incidents at AI developer platforms this month, including Railway, OpenCode, Warp and Lightning AI. In two cases Wiz investigated, stolen enterprise credentials ran about 2.5 million Amazon Bedrock API calls, roughly $100,000 of AI usage. Wiz has not said the incidents share one attacker.
The pattern matters for small teams. A leaked database password gets you a data breach. A leaked AI key gets you a bill, charged to whoever owns the account. If your startup keeps provider keys in a hosting platform’s environment variables, treat each one as money in a drawer. Give every service its own key, cap each key’s spend, and know who gets the billing alert. The same lesson ran through the backdoored Claude Code extension in Cursor: developer tools now hold the keys to paid AI accounts, so they need the same care as a payment card. The same thinking is behind 1Password for Agents, which fills logins for an AI agent without ever showing it the password.