Northflank Tells Customers to Rotate Secrets After AI API Key Abuse
Cloud application hosting provider Northflank warned customers on October 30 to rotate credentials and secrets stored on its platform after a breach led to unauthorized use of customer AI API keys. The company said customers should assume that environment variables had been decrypted and stolen, creating a risk of unauthorized access and unexpected AI charges.
Environment variables are commonly used for database passwords, cloud credentials, authentication tokens and API keys. Northflank later confirmed reports that AI keys supplied to the platform had been used without authorization. A stolen key may allow access to connected services, data theft or charges made through the affected provider.
The incident followed an alert from cloud security company Wiz at 11:43 UTC on October 30. Northflank said its engineers reviewed the alert one minute later and applied a network policy patch, followed by additional patches at 12:50 UTC and 13:05 UTC. The company also requested help from Wiz's incident response team.
Northflank has not disclosed how the attackers obtained access, how many customers were affected or whether credentials other than the reported AI API keys were stolen or abused. That limits what customers can determine about the scope of the incident from the company's current disclosures.
Customers should rotate secrets stored in workloads, secret groups and global secrets, and review AI provider usage and billing logs for unexpected activity. Northflank also advised reporting suspicious use to the company and the relevant AI providers. It enabled credential rotation for PostgreSQL, MongoDB, MySQL and Redis, and turned on audit logs for all customers, including free accounts.
Wiz has described stolen AI credentials as a route to so-called LLMJacking, in which attackers use another organization's access to AI models. Northflank said customer reports may be connected to attacks on AI development platforms documented by Wiz, but Wiz has not confirmed that all of the incidents it reviewed were carried out by the same attackers.