--- title: "Northflank Breach Led to Abuse of Customers' AI API Keys" description: "Northflank tells every customer to assume their environment variables were decrypted and to rotate all secrets, after users reported AI API keys being misused." url: https://publishmyblog.com/website-tools/northflank-breach-ai-api-keys/ section: "Website Tools" author: "Saeed Shahriar" published: 2026-11-01T20:33:38.220Z --- # Northflank Breach Led to Abuse of Customers' AI API Keys If your app keeps an OpenAI or Anthropic key in a hosting platform's env vars, this is the month to check your bill. Northflank, the cloud platform developers use to deploy apps, databases and AI workloads, says **every environment variable on it should be treated as stolen**. Some customers have already seen their AI API keys used by someone else, so if your app keeps an OpenAI or Anthropic key there, assume someone else has it too. Do you host anything there? Rotate every secret today. ## What Northflank has said so far The alert came from Wiz at 11:43 UTC on October 30, says the timeline on [Northflank's incident page](https://status.northflank.com/). **A first network policy patch went out at 12:02 UTC.** Engineers looked at the alert within a minute, and more patches followed at 12:50 and 13:05 while Northflank asked Wiz's incident response team to help. By 13:58 UTC the advice had hardened: "Assume all environment variables have been decrypted." Then came the bill. **Multiple users confirmed their AI API keys had been used.** Late on October 30, Northflank said users had reported unauthorised use of AI API keys supplied to the platform, and it told everyone to rotate their secrets immediately. Its four asks for you: - Revoke each credential with the provider that issued it, and make a new one. - Update your affected secrets and workloads in Northflank. - Check your AI provider's usage page and billing logs for anything you don't recognise. - Report suspicious usage to Northflank support and to that provider. It also says to switch off public access on add-ons where you'd turned it on, and to keep going until every workload, secret group and global secret has a fresh value. ## What's still unknown Quite a lot, actually. As [BleepingComputer reports](https://www.bleepingcomputer.com/news/security/northflank-security-breach-leads-to-abuse-of-customers-ai-api-keys/), Northflank hasn't said: - how attackers got in - how many customers were hit - whether credentials other than the AI API keys were abused BleepingComputer asked. No reply by publication. Meanwhile, the company has shipped tools. **Secret rotation now works for six add-on types**: Postgres, MongoDB, MySQL, Redis, Minio and RabbitMQ. On October 31 it turned on **audit logs for every customer**, free accounts included. ## Part of a bigger October Wiz's report, ["Black October" for AI Developer Platforms](https://www.wiz.io/blog/black-october-for-ai-developer-platforms), lists nine platforms that disclosed incidents this month. **Northflank is the ninth**, after Railway, OpenCode, Warp and others. Wiz says every public report so far shows attackers hunting for stored environment variables, with **a focus on the ones that grant AI access**. Northflank says its users' reports may fit that pattern. ## Why attackers want AI keys Still, why steal AI API keys at all? **Because stolen keys are easy to sell.** Wiz calls it LLMJacking: running AI models on stolen AI API keys, often for resale. In two cases Wiz traced to one of these vendor breaches (it doesn't say which), **stolen keys made about 2.5 million Amazon Bedrock calls**, worth over $100k at list price. - **The open key:** it allowed calls through any region, so only about 6% were throttled. - **The scoped key:** limited to US regions, it kept running into regional quotas. In both cases, the first test call on the stolen key came **within 30 minutes of the reported credential access**, from a residential IP address. The flood came days later. Speed matters here. ## Why it matters Run a SaaS or an AI feature? Your **OpenAI or Anthropic key probably sits in an environment variable** on a platform you don't control, next to your database password. So that platform's security is now part of your AI bill. My guess is most of you won't spot it until the usage page or the invoice shows a spike, and by then a stolen key may have been busy for days. I've covered [the DebugBear script hijack](https://publishmyblog.com/website-tools/debugbear-script-hijack-polymarket/) and [the UpdraftPlus breach](https://publishmyblog.com/website-tools/updraftplus-updraftcentral-breach/) this month, and in both a vendor you trusted became the way in. So these are the checks I'd run this week, even if you've never used Northflank: 1. Open your AI provider's usage page and read the last two weeks, day by day. 2. Give each app its own key. 3. Set spending limits or budget alerts wherever your provider offers them, because a stolen key with no cap can run all night before you notice. 4. Scope your keys tightly, the way Wiz's second case shows. Rotation is the step people skip. **I think most teams will rotate the AI key and forget the database password next to it.** Northflank said rotate every secret. That means the database passwords too. And if you're not on Northflank, check your AI provider's usage page anyway. ## Common Questions **What should Northflank customers do after the October 2026 security incident?** Rotate every secret stored in Northflank, including workload secrets, secret groups and global secrets. Revoke old keys with each provider, check your AI provider's usage page and billing logs, and report odd usage to Northflank support. **Were Northflank customers' AI API keys actually misused?** Yes. Northflank said on October 30, 2026 that users reported unauthorised use of AI API keys supplied to its platform. If you used Northflank, check your AI provider's usage page for calls you didn't make. **What is LLMJacking and how does it relate to the Northflank breach?** LLMJacking means using stolen credentials to run AI models on another organisation's account. Wiz lists Northflank among nine October 2026 platform incidents and says attackers hunted stored environment variables, especially ones holding AI API keys. ## Sources - [Northflank security breach leads to abuse of customers' AI API keys](https://www.bleepingcomputer.com/news/security/northflank-security-breach-leads-to-abuse-of-customers-ai-api-keys/), BleepingComputer - [Security incident](https://status.northflank.com/), Northflank - [“Black October” for AI Developer Platforms](https://www.wiz.io/blog/black-october-for-ai-developer-platforms), Wiz