Northflank warns customers after AI API key abuse

Northflank is a cloud platform for deploying and managing applications, containers, databases and AI applications without managing the underlying infrastructure.[1] It disclosed the incident on October 30 after receiving a threat alert from Wiz, a cloud security company.[1] Northflank told customers to assume their environment variables had been decrypted and to disable or rotate all keys.[2] The company said multiple customers had reported unauthorized use of AI API keys supplied to its platform.[2] Northflank advised customers to rotate secrets and credentials immediately, including those in workloads, secret groups and global secrets.[2] It also urged customers to revoke existing credentials with the providers that issued them, replace them, and update affected secrets and workloads.[2]

Northflank advised customers to review AI-provider usage and billing logs for unexpected activity and to report suspicious use to Northflank and the relevant provider.[2] It also recommended disabling public access to add-ons where that access was enabled.[2] The company enabled audit logs for all customers, including free users.[2] It made secret rotation available to all users for PostgreSQL and MongoDB, and to all customers for MySQL and Redis.[2] Rotation was also made available for Minio and RabbitMQ add-ons.[2] These changes give customers specific controls to use while replacing credentials and reviewing affected workloads.

BleepingComputer reported that Northflank had not said how many customers were affected or how attackers gained access.[4] Northflank says environment variables commonly store sensitive information, including database passwords, cloud credentials, authentication tokens and API keys.[1] BleepingComputer reported that exposed credentials could be used to breach other customer services, potentially resulting in data theft or unauthorized API charges.[1] Wiz reported that public reports across October incidents highlighted targeted searches for stored environment variables, especially those that grant AI access.[3] Wiz said it found no single vulnerability or dependency linking the incidents.[3] The available information therefore does not establish how the Northflank incident began or whether it shared a cause with other reported incidents.

Wiz reported that at least nine AI developer platforms disclosed incidents during October, including Northflank and platforms such as Fireworks AI, Railway and Warp.[3] In two enterprise cases, Wiz said likely usage at list price exceeded $100,000.[3] It observed about 800,000 Amazon Bedrock calls over 30 hours in one case, with about 90% succeeding; in another, it observed about 1.7 million calls over 13 hours, with more than one million succeeding.[3] Wiz said the first unauthorized activity in both cases occurred within 30 minutes of reported credential access at the breached vendor.[3] Wiz described LLMJacking as a growing threat linked to a market for LLM brokers that can make stolen AI credentials monetizable.[3]