Skip to content

SecurityRailway

Under investigation

Railway incident report: attackers hunted AI provider credentials in the October breach

Railway platform at a glance

Railway's incident report, published November 10, 2026, says the October attacker decrypted environment variables for at most 1.85% of workspaces (about 100,000), exposed linked tokens for 1,852 users, and ran up $740,000 in unauthorized AI spend on stolen AI provider credentials, which Railway will backstop.

Railway platform: status

Under investigationNot rated
Affected
  • Railway workspaces (at most 1.85%, about 100,000, as an upper bound): a subset of customer environment variables was decrypted, including AI provider credentials
  • 1,852 Railway users: linked-account tokens (BYOK agent keys and GitHub OAuth tokens) were in the decrypted data; Railway reset all of them
  • 353 Railway builds: stored build histories were read, sealed variables included
  • Railway says the customer database was not accessed and no Railway account logins used credentials taken in the incident
CVE identifiers
Not published
Last checked

What changed with Railway platform

Railway published its final incident report on November 10, 2026, covering the October 2026 attack it first disclosed on October 21, in which the attacker hunted AI provider credentials stored in customer environment variables. The attacker probed Railway's build infrastructure from October 18, first exploited it at 07:28 UTC on October 20, chained exploits to move laterally across systems, and was fully evicted at 09:24 UTC on October 21. Railway halted builds twice, rotated internal credentials, invalidated every user session and token, and restored builds for all plans by October 22 after hardening the infrastructure.

The report puts the customer impact at an upper bound: decrypted environment variables for at most 1.85% of workspaces, about 100,000, with the decryption of per-environment keys running from 04:42 to 07:14 UTC on October 21. The attacker also read the stored histories of 353 builds, sealed variables included, and the decrypted data included linked-account tokens, BYOK agent keys and GitHub OAuth tokens, for 1,852 users; Railway reset every one of those tokens and says it identified no logins to Railway accounts using credentials taken in the incident, and that the customer database was not accessed.

Railway's forensic investigation concluded the attacker was specifically targeting AI provider credentials to sell to distributed transfer stations for use in distillation, using safeguard-stripped models with a harness similar to ARTEX, according to Wiz. Unauthorized spend on the stolen keys was estimated at about $740,000; Railway is coordinating refunds with providers and will cover the losses itself where a provider will not. Railway says every attack vector used has been remediated: builds now run on dedicated ephemeral virtual machines, the CI pipeline was rebuilt, and secret storage and decryption were hardened. Its specialist forensics firm's investigation is still ongoing, and indicators of compromise will be shared publicly once law enforcement signs off.

How to fix it: Railway platform

  1. 1. Configuration

    Rotate all environment variables stored in Railway, including any API keys, as well as private registry credentials, and review logs for those services for unusual activity, per Railway's October 21 disclosure.

    blog.railway.com (opens in a new tab)
  2. 2. Monitor

    Report AI charges you did not make through Railway's unauthorized-spend form: Railway is coordinating provider refunds and says it will cover token losses itself where a provider will not.

    blog.railway.com (opens in a new tab)

Advisories

Railway platform in numbers

Values are copied from the linked source on the date shown. They are not IA Compa measurements.

Unauthorized AI spend on stolen keys (Railway estimate)as of · Source for Unauthorized AI spend on stolen keys (Railway estimate): blog.railway.com (opens in a new tab)
$740,000

Why Railway platform matters for businesses

Why it matters: if your agency or clients deploy on Railway, or store AI keys in any host's environment variables, this is a rotation event, not a headline. Rotate every AI provider key that sat in a Railway workspace before October 21, reconnect GitHub, and check provider bills for October usage you did not make, then claim through Railway's form. The wider lesson of the October wave, after Fireworks and OpenCode, is that platforms holding many customers' AI keys are now a primary target, because one breach turns into spendable inference.

Get AI News by email

Verified AI stories, each with its official source, on Monday and Thursday mornings. Add breaking news to hear about the biggest ones as soon as we publish them.

Every story links its official source

Free, in English or Spanish

Unsubscribe with one click, any time

What to send

Sources for this story

  1. Official
  2. Official

Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.

  • #security
  • #data-breach
  • #ai-provider-credentials
  • #api-keys
  • #environment-variables
  • #railway-incident-report
  • #incident-response

Want this working in your business?

We turn new AI capabilities into agents, voice receptionists, and automations you can actually use.