At least nine AI developer platforms were breached in October, and attackers are going after the AI API keys stored on them.

Security firm Wiz counted at least nine AI developer and hosting platforms that disclosed security incidents in October, and it calls the wave "Black October". The list: Fireworks AI, Andromeda, Zeabur, Lightning AI, OpenCode, Railway, Warp, CommandCode and, on Oct 30, Northflank. The way in differed, but the target kept repeating: the stored settings (environment variables) where developers keep the API keys that unlock paid AI models. Stolen keys get used to run AI models on the victim's bill or are resold, which is known as LLMjacking. In two cases Wiz looked at, stolen cloud credentials ran about 800,000 AI model calls in 30 hours and about 1.7 million in 13 hours, likely costing more than $100,000 at list prices. Wiz saw the first test call within 30 minutes of a credential being exposed, and it says it can't confirm whether one group is behind all of these. Zeabur is shutting down after its incident. Northflank told all customers to "assume all environment variables have been decrypted" and to rotate every key, and later said users had reported unauthorised use of AI API keys they'd stored on the platform (Northflank status page, BleepingComputer).

What this means for you: if you've put an OpenAI, Anthropic, Google, AWS or other AI key into one of these platforms, cancel it with the company that issued it and make a new one. Changing it only on the hosting platform isn't enough. Then check that provider's usage and billing pages for calls you didn't make. Even if you don't use these platforms, setting a spending limit on your AI accounts caps the damage if a key ever leaks.