العربية
🚨 BreakingFBI Warns Artex AI Tool Automates Hacks at Machine Speed

Railway Breach: Stolen AI Keys Ran Up $740,000 in Charges

Security World Teamاقرأ بالعربية

Hosting platform Railway said AI provider keys stored by customers were stolen in an October breach and used for an estimated $740,000 in unauthorized spending. It will pay the charges itself when a provider refuses a refund.

Illustration for Cybersecurity News
Illustration by Security World

⚡ Why it matters

AI keys are a lucrative target because attackers burn your credit directly. Any secret stored as an environment variable with a hosting provider is only as safe as that provider.

Railway hosts and deploys customer apps and keeps their secrets, for example API keys and database passwords, in environment variables.

What happened

  • Variables were decrypted in up to 1.85% of workspaces (roughly 100,000 at most), and the attacker's queries hunted for AI provider keys.
  • Decrypted data included linked-account tokens for 1,852 users, such as BYOK agent keys and GitHub OAuth tokens. Railway reset every one, and found no sign-ins made with them.
  • One query also matched cloud, database and payment credentials, but forensics showed no access to customer databases or repository downloads.
  • The attacker read stored histories of 353 builds, including the secrets passed to them.

A change in stance

On October 26, Railway said it could not refund spend made with stolen keys. It now says it will cover charges providers decline to refund. Provider policies differ: AWS Bedrock and Google Gemini have started refunds, OpenAI and SpaceXSI review cases individually, and talks with Anthropic are ongoing.

What to do

  1. Rotate every AI key and credential stored on Railway, even if you were not emailed.
  2. Set spending caps and usage alerts with your AI providers.
  3. Separate production and development keys, and give each key the least privilege possible.

🔗 Sources

This article is original reporting by the Security World editorial team based on the sources listed above; direct quotes are kept short and marked with quotation marks. Corrections and feedback: info [at] securitywd.net.

Related stories