Northflank Security Breach Exposes AI API Keys, Prompts Widespread Secret Rotation

Cloud-native platform Northflank is currently investigating a security incident that has led to the reported misuse of AI API keys. In response, the company has issued an urgent directive to its customers, advising them to rotate all platform secrets immediately. The breach may have resulted in the exposure of sensitive environment variables across various workloads, secret groups, and global settings.

While Northflank is actively investigating the incident, the exact entry point for the attackers, their identity, and the full scope of the compromise remain unclear. The company has not yet confirmed widespread theft but is treating the situation with extreme caution. Northflank's published timeline indicates that the incident came to light on October 30, 2026, following a threat alert from Wiz at 11:43 UTC. Engineers responded rapidly, deploying an initial network policy patch within minutes, followed by further patches later that day. However, the details of these patches and how they address the initial intrusion vector are not yet public.

Following the initial alerts and patching efforts, Northflank instructed users to consider all environment variables as potentially compromised. They also advised restricting public access to any enabled add-ons. Environment variables are a common storage location for critical credentials such as API keys, database passwords, and other sensitive application secrets, making their exposure a significant security risk.

This incident bears similarities to a recent security breach at Railway, where customers were also urged to replace stored secrets following potential unauthorized access. Northflank suggests that the reported activity aligns with a broader trend of attacks targeting AI credentials. The company pointed to similar IP addresses identified in Wiz's "Black October" report, which detailed a series of incidents affecting nine AI developer platforms throughout October. Wiz noted a common focus on stored environment variables across these attacks but found no single vulnerability connecting them.

The primary financial risk associated with such breaches is "LLMjacking," a scenario where attackers exploit stolen credentials to consume paid AI services at the victim's expense. Previous analyses of LLMjacking attacks have demonstrated how compromised cloud keys can facilitate unauthorized access to and resale of AI models. However, Northflank has not yet disclosed any confirmed financial losses or linked its incident to a specific resale operation.

Northflank is guiding customers through a comprehensive remediation process. This includes revoking existing credentials directly through their issuing providers, generating new replacement credentials, and subsequently updating all affected Northflank secrets and workloads. Furthermore, customers are advised to meticulously review their AI provider usage and billing records for any unexpected activity. Any suspicious findings should be reported to Northflank's support team and the relevant AI service provider.

Secret rotation capabilities have been extended to cover a range of add-ons, including PostgreSQL, MongoDB, MySQL, Redis, MinIO, and RabbitMQ. In a move to enhance transparency and aid investigations, Northflank has also enabled audit logs for all users, including those on free tiers. These logs can provide crucial records for identifying the scope and nature of the compromise.

The immediate priority for Northflank and its customers is the replacement of all potentially exposed credentials, not solely those AI keys already confirmed to be abused. The company's ongoing investigation continues to cover secrets across workloads, shared secret groups, and global settings, emphasizing a thorough approach to securing its platform and user data.