THE PERMISSION [LAYER]

AI agent incident

Railway says an intruder decrypted customer variables to hunt AI provider keys, with about $740,000 in unauthorized AI spend

Reported · 3 sources

  • credential-theft
  • spend
  • autonomous-attack

What happened

Railway, a hosting platform, published its full incident report on November 10 on an October intrusion. Starting at 07:28 UTC on October 20 an attacker partly escaped filesystem isolation on a build host, chained further exploits across internal systems, and decrypted customer variables for at most 1.85% of workspaces, about 100,000, before being expelled at 09:24 UTC on October 21. Railway says the attacker was specifically targeting AI provider credentials, appeared to use models with their safeguards removed, and that unauthorized spend on the stolen keys was estimated to reach about $740,000. Each build now runs in its own short-lived virtual machine with a credential that can only push one image.

Why it matters

A model key stored as an environment variable on a hosting platform is a spending credential held by a third party. Wiz lists Railway among 16 AI developer platforms that disclosed incidents in the same weeks. Rotate any key stored there, and set a spend cap and a network restriction at the provider so a copied key cannot run unbounded.

New incidents by email

Get incidents like this one in your inbox

New AI agent incidents are covered in the briefing, with what to check on the agents you already run. At least once a week, free.

You confirm by email. Unsubscribe in one click.

Sources

  1. blog.railway.com Incident report: October security incident
  2. wiz.io "Black October" for AI Developer Platforms
  3. cloud.google.com The Rise of LLMJacking: How Threat Actors Monetize Compromised Cloud and Model Resources

Background reading

What is an LLM gateway? An explainer on the problem behind this incident.

Covered in the briefing

Issue #4: A wave of AI platform breaches is after model keys November 10, 2026

Cite or correct this record

“Railway says an intruder decrypted customer variables to hunt AI provider keys, with about $740,000 in unauthorized AI spend.” The Permission Layer AI Agent Incident Database, https://thepermissionlayer.com/incidents/railway-build-breach-ai-provider-keys/. Database updated November 10, 2026.

Every record traces to the dated sources above. If something here is wrong or missing, email editor@thepermissionlayer.com and it gets fixed. See this record in the full list.

The Permission Layer

Get the next incident like this explained

When an AI agent breaks out, leaks data or runs up a bill, the briefing covers what happened and what to check on the agents you already run. At least once a week, five-minute read, free.

You confirm by email. Unsubscribe in one click.