Reported by 2 outlets
The story so far
AI summary of the reporting below · Updated · How we work
The intrusion started on October 20 at approximately 07:28 UTC and lasted roughly 26 hours until defenders contained it on October 21 at 09:24 UTC. Railway estimated that decrypted environment variables across at most 100,000 workspaces—about 1.85% of its total—were exposed, alongside linked-account tokens belonging to 1,852 users. [1][2]
The threat actors initiated the breach by exploiting a zero-day vulnerability to escape filesystem isolation on multi-tenant build machines and steal an internal credential. They used that access to inspect Railway's internal source code, uncover a misconfigured authentication mechanism, and obtain keys needed to decrypt customer environment variables. The attackers targeted AI API keys to drive coding agents and model distillation pipelines, while Railway engineers engaged in a rapid, parallel defensive effort against automated AI agents before rotating credentials and patching the flaws. [1][2]
Railway stated that it found no evidence of attackers downloading customer Git repositories or accessing hosted databases. The company is collaborating with AI providers to secure reimbursements for impacted customers and promised to cover any fraudulent balances that providers decline to refund. Security firm Wiz linked the intrusion to "Black October," an ongoing wave of automated attacks aimed at developers and AI platforms including Northflank, Warp, OpenCode, and Lightning AI. [1][2]
Cloud hosting provider Railway says hackers exploited vulnerabilities in its infrastructure to steal customer authentication secrets and generate approximately $740,000 in fraudulent AI usage, potentially exposing credentials for up to 100,000 workspaces.
via BleepingComputer
A zero-day in a cloud provider's infrastructure granted attackers access to paid AI models and in a single day they raked up hundreds of thousands in expenses.
via TechRadar