Northflank customers should replace credentials stored on the cloud hosting platform after a breach prompted reports of unauthorized AI service use. BleepingComputer reported on October 31 that Northflank had disclosed the incident the previous day and instructed all customers to treat secrets in their environment variables as stolen.
Northflank hosts applications, containers, databases and AI applications, handling the underlying infrastructure for developers. Its warning reaches beyond AI keys: customers are being told to replace all stored credentials, although the abuse reported by multiple customers involves AI API keys—credentials that authorize access to AI services.
Environment variables supply settings to running applications and often hold sensitive values, such as database passwords, cloud service credentials and authentication tokens. A stolen database password, for example, could give an attacker access to another service used by a Northflank customer. Exposed API keys can also leave customers paying for someone else’s usage.
What Northflank has reported
The response began with an alert from cloud security company Wiz. Northflank’s published timeline puts that warning at 11:43 UTC on October 30, with engineers reviewing it one minute later. A patch to network policy came first, followed by further patches at 12:50 UTC and 13:05 UTC. Northflank also sought help from Wiz’s incident response team.
The company initially called for immediate rotation of environment variables and API keys. An October 30 update then told customers to assume all environment variables had been decrypted and to disable or replace every key. Northflank subsequently confirmed that customers had reported unauthorized use of AI keys they had supplied to the platform.
The breadth of that precautionary instruction exceeds the reported abuse: BleepingComputer’s October 31 account described an undisclosed entry method and customer total, with theft or misuse of other credential types also unresolved. Northflank had not answered the publication’s questions about access to stored credentials and the impact on customers by publication time.
Why attackers seek AI credentials
Wiz placed Northflank among at least nine incidents affecting AI development platforms during October in its report, “Black October for AI Developer Platforms.” Other platforms named included Railway, OpenCode, Warp and Lightning AI. The incidents included potentially exposed decrypted environment variables at Railway, exposed AI API keys at OpenCode, and exposure of AI agent logs and user telemetry at Warp.
Researchers view these platforms as attractive targets because their stored keys can unlock expensive AI services. Wiz said attackers in some of the incidents specifically sought environment variables holding AI credentials.
That access enables LLMJacking: using stolen credentials to run AI models at the credential owner’s expense, or reselling the access. Across two cases Wiz investigated, compromised enterprise credentials powered approximately 2.5 million Amazon Bedrock API calls, generating approximately $100,000 in AI usage.
Wiz also found infrastructure associated with AI access resale services, though it could not connect those operations to the initial breaches or the large-scale abuse. Many incidents involved attempts to steal AI keys, but Wiz has not established that one attacker was responsible for all nine. Northflank cited the research in its updates, suggesting that its customers’ reports may relate to the activity Wiz observed.
Replace secrets and check usage
Northflank’s instructions cover secrets held in workloads, secret groups and global secrets. Customers should rotate those values and disable public access to add-ons wherever it is enabled.
To support the response, Northflank enabled credential rotation for PostgreSQL, MongoDB, MySQL and Redis. It also enabled audit logs across all customer accounts, including the free tier, so customers can investigate suspicious activity.
Customers should check their AI providers’ usage and billing logs for unexpected activity and report suspicious usage to both Northflank and the relevant AI provider.
