Nine AI developer platforms disclosed credential breaches during October 2026, in a run cloud security firm Wiz has called "Black October". The platforms named are Fireworks AI (6 October), Andromeda (9 October), Zeabur (14 October), Lightning AI (15 October), OpenCode (16–17 October), Railway (20 October), Warp (23 October), CommandCode (28 October) and Northflank (30 October).

What the attackers went after

In every case, the public reports describe attackers hunting stored environment variables — the credentials that unlock AI models and cloud accounts — held by intermediary platforms on behalf of many customers at once. Fireworks AI, OpenCode and Warp reported signs of an "agent swarm" attack, while OpenCode and CommandCode attributed their incidents to distillation attempts, in which stolen model access is used to train or copy models. Root causes differ across the nine, and no single actor behind all of them has been publicly claimed.

How fast stolen keys were abused

Wiz traced two enterprise keys to Amazon Bedrock abuse worth more than $100,000 at list price. In both cases the first unauthorised call came within 30 minutes of the reported credential access, followed days later by floods of about 800,000 calls in 30 hours and 1.7 million calls in 13 hours. Stolen access was then resold as coding-agent usage. Wiz's advice to customers of the affected platforms is blunt: rotate stored keys now, scope keys per project, and watch for unexpected spend spikes, because a stolen key looks like normal traffic until the bill arrives.