Northflank, a cloud platform developers use to deploy applications, databases and AI workloads, has told all of its customers to assume the secrets they stored with it were decrypted and to replace every key, BleepingComputer reported Saturday. Northflank is at least the ninth AI developer platform to disclose a security incident in October, according to security firm Wiz, a pattern that turns one vendor's breach into its customers' bill.
What Happened
According to its public incident timeline, Northflank began responding Friday within a minute of a threat alert from Wiz, patched its network policies and asked Wiz's incident response team for help. The same day it told customers to treat all environment variables, the settings where applications keep passwords and API keys, as decrypted, and to disable or rotate every key in their workloads and shared secret stores.
Later Friday, multiple users confirmed unauthorized use of AI API keys they had supplied to the platform. Northflank told them to revoke those keys with their AI providers and check billing logs, linking the reports to a broader pattern of attacks on AI credentials. It has since switched on audit logs for every customer, including free accounts, and added credential rotation for its managed PostgreSQL, MongoDB, MySQL and Redis databases.
Northflank has not said how attackers got in or how many customers were affected, and it had not answered BleepingComputer's questions at publication.
Why It Matters
In a report titled "Black October for AI Developer Platforms," published Friday, Wiz lists nine platforms that disclosed incidents this month, including Fireworks AI, Lightning AI, OpenCode, Railway, Warp and Northflank. Wiz says public reports describe attackers hunting for stored environment variables, especially those that unlock AI models. The practice is known as LLMjacking: thieves run or resell model access at the victim's expense. One platform, Zeabur, is shutting down because of its incident, Wiz says.
In two cases it attributes to vendor breaches, stolen corporate keys for Amazon Bedrock, Amazon's AI model service, were tested within 30 minutes of the reported theft. Days later, one key made about 800,000 calls in roughly 30 hours and the other about 1.7 million in roughly 13 hours. Wiz, which sells cloud security products, estimates the two cases alone ran up more than $100,000 at list price.
For boards, an AI key is now a payment instrument, and every key handed to a third-party platform carries billing exposure. Useful controls include provider spending caps, a separate key for each project and vendor, and alerts on unusual usage.
Context
The Primary Wire reported the OpenCode breach Oct. 19 and Railway's warning to rotate secrets Oct. 22, both citing BleepingComputer. Wiz says it has not confirmed that the same attackers are behind all nine incidents. It tied two resellers of cheap AI model access to testing of the stolen Bedrock keys but could not link them to the break-ins.
What to Watch
- Whether Northflank publishes a root cause and a count of affected customers.
- Whether Wiz or the affected platforms connect the October incidents to one group.
- Further disclosures from AI developer platforms.
