Northflank Security Breach Leads to AI API Key Abuse

Cloud application hosting platform Northflank has confirmed a security breach that led to the unauthorised use of customers’ artificial intelligence (AI) API keys. The company is urging all customers to immediately rotate credentials and secrets stored on its platform and review their AI provider usage logs for suspicious activity.

The incident followed a warning Northflank received from cloud security company Wiz on October 30. Northflank quickly responded with network policy patches and further security enhancements, also requesting assistance from Wiz’s incident response team.

Initially, Northflank advised customers to assume all environment variables on the platform had been decrypted and the secrets within them stolen. Environment variables are commonly used by applications to store sensitive data such as database passwords, cloud service credentials, authentication tokens, and API keys. Their exposure can lead to further compromises of customer services, including data theft or unauthorised billing.

Northflank later confirmed that multiple customers had reported suspicious activity, specifically the unauthorised use of AI API keys. "Users have reported unauthorised use of AI API keys supplied to the Northflank platform," the company stated in an October 30 update.

The company has not disclosed how attackers gained access to its systems, the total number of affected customers, or whether credentials other than AI API keys were stolen or abused. As part of its remediation, Northflank enabled credential rotation for several database services, including PostgreSQL, MongoDB, MySQL, and Redis. It also activated audit logs for all customers, including free accounts, to aid investigations. Customers are further advised to disable public access to add-ons where enabled, and to rotate secrets stored in workloads, secret groups, and global secrets.

Wider Trend of AI Platform Attacks

The Northflank breach occurred amidst a wider series of attacks targeting AI development platforms, which Wiz has described in a report titled "Black October for AI Developer Platforms." Wiz identified at least nine security incidents affecting such platforms during the month, including Railway, OpenCode, Warp, and Lightning AI. Many of these incidents involved attempts to steal AI API keys stored in environment variables.

Wiz researchers highlighted a technique they call "LLMJacking," where stolen credentials are used to access AI models via another organisation’s API keys. This can lead to significant unauthorised charges or allow stolen access to be resold. In two cases investigated by Wiz, compromised enterprise credentials were used to make approximately 2.5 million Amazon Bedrock API calls, generating an estimated $100,000 in AI usage costs.

While Wiz has not confirmed that all nine incidents were carried out by the same attackers, the common targeting of AI API keys suggests a significant trend. Northflank itself noted that customer reports of unauthorised AI key usage may be connected to the attacks observed by Wiz.