TLDR
- On October 21, 2026, Railway disclosed unauthorized access to customer environment variables, followed by a preliminary report on October 22. The attack occurred from October 20, 13:58 to October 22, 03:32 UTC.
- Scope: Environment variables were potentially compromised in up to 1.85% of workspaces, around 100,000. However, Railway advises everyone to rotate secrets stored before October 21.
- The attack targeted AI provider keys for distillation attacks. Railway requested GitHub, GitLab, Anthropic, OpenAI, OpenRouter, Stripe, and others to revoke keys, and in some cases, provided them with all keys from the platform. Receiving a key revocation notice doesn't necessarily mean your specific key was leaked.
- User accounts were not compromised. Railway will confirm access to source code in the final report.
- Builds and deployments for all plans were restored on the evening of October 22. Railway also reset OAuth sessions, rotated storage bucket credentials, and revoked all API tokens.
- Independently, researcher Taylor Monahan claimed the theft of approximately $371,000 in crypto from wallets whose private keys were stored on Railway. Railway has not confirmed this information.
Updated October 24, 2026. The investigation is ongoing, and Railway promises to publish the final report later. The current status is available at status.railway.com.
Railway is a cloud platform for deploying applications. The incident affects not only the platform itself: if keys to external services were stored in Railway's environment variables, those services are also at risk. Below is what is known so far, what the company says, and what users should do.
🎯 Section 1. What Happened
According to Railway's preliminary report, the attack began on October 20 and lasted until 03:32 UTC on October 22, when the company confirmed the attacker no longer had access to internal systems.
The attack path, in general terms, was as follows: privilege escalation in the build infrastructure → access to CI environment credentials → decryption of artifacts with additional API access → unauthorized request to encrypted environment variables, some of which were successfully decrypted. Railway has not yet disclosed technical details of the exploits.
Due to envelope encryption, the company cannot determine which specific variables in affected projects were decrypted. Therefore, the recommendation to rotate applies to all secrets, not just individual keys.
Developers typically store the most sensitive data in environment variables: API keys, database passwords, and credentials for private container registries. Thus, not only the platform but also all external services whose keys were stored on it are at risk.
Independent researchers have reported potential financial consequences. Taylor Monahan, according to The Hacker News, linked the incident to the leak of crypto wallets whose keys were on Railway; CryptoSlate estimates the amount at approximately $371,000 and notes that Railway has not confirmed this theft or its connection to the incident. A specialist from Symantec Threat Hunter Team, however, told The Hacker News that they saw no evidence linking the activity they observed to the Railway incident.
🎯 Section 2. Incident Timeline
All times are UTC, according to Railway's preliminary report and status page.
- October 20, 13:58 — Monitoring detects anomalous activity. By 16:51, Railway identifies a privilege escalation attempt in the build infrastructure and halts build pipelines.
- October 20, 19:53 — Evidence emerges that the attacker may be able to expand access in several parts of the system.
- October 21, 04:24–07:25 — The attacker exfiltrates secrets in encrypted form and later obtains CI environment credentials.
- October 21, 07:25–09:18 — Railway discovers an unauthorized request to encrypted environment variables; some are decrypted.
- October 21, 09:25 — Railway requests providers to revoke keys.
- October 21, 20:19–22:10 — Session invalidation, security disclosure publication, and rotation recommendations.
- October 22, 03:32 — Confirmed that the attacker no longer has access. Builds for Enterprise and Pro are restored by 06:29; all Railway API tokens are revoked at 04:17.
- October 22, 17:00–22:49 — API and deployments for all plans, including Hobby, Free, and Trial, are fully restored.
- October 22, evening — Railway publishes a preliminary report.
- October 23 — Railway sends separate emails to workspace owners at higher risk.
🎯 Section 3. What Railway Says and What It Has Already Done
According to Railway, the attacker was searching for AI provider and platform keys to use them for model distillation. The company considers the following to be at the highest risk:
- workspaces with a very large number of environment variables;
- environments with a large number of AI keys from various providers.
Environment variables were potentially compromised in up to 1.85% of workspaces, approximately 100,000. However, Railway provided providers with a much broader list: the company shared all keys from the platform with some of them, regardless of whether they were compromised. Therefore, receiving a key revocation notice from OpenRouter, GitHub, or Stripe does not necessarily mean your specific key was leaked.
The attacker did not compromise user accounts. Railway has not yet ruled out access to source code; this will be confirmed in the final report.
What Railway has already done:
- paused builds and deployments, and after restoration, verified that the same escalation patterns no longer work;
- reset OAuth sessions, revoked all API tokens and Project Tokens created before the incident — new ones are required for CLI and Public API;
- rotated credentials for all storage buckets (account logs might show a revocation notice from a Railway employee — according to the company, this is expected);
- reissued all internal keys twice, rebuilt infrastructure tools from verified code;
- expedited the transition to a new generation of builders — isolated single-tenant environments instead of multi-tenant;
- reissued all Railway GitHub App installations;
- engaged an external company for forensics.
Railway has also published guides on checking key usage and rotating credentials, and promises the final report after the investigation is complete.
🎯 Section 4. Why Deployments Were Paused
The entry point for the attack was the build infrastructure: Railway detected privilege escalation there and halted build pipelines on October 20, and again on October 21 when new access routes were found.
The second reason is the order of operations. Railway deliberately first removed the attacker from the infrastructure and only then asked users to rotate keys. Otherwise, new credentials could have been compromised immediately after they were entered.
Already running services continued to operate throughout this period. Deployments were restored in stages: on the morning of October 22 for Enterprise and Pro, and that same evening for all plans.
🎯 Section 5. How This Manifests in Your Projects
If your application on Railway interacts with AI APIs or external services, you might have observed:
401 Unauthorized (or 402 Payment Required) errors from AI providers, even with a positive balance and no code changes — providers preemptively revoked keys at Railway's request;
- New values in Variables not being applied: the container continues to use old variables until redeployed;
- Failures in CLI and Public API integrations: all API tokens and Project Tokens created before the incident have been revoked, new ones are needed;
- Issues with storage buckets: new credentials appear in the dashboard, but the service needs a redeploy to pick them up;
- Projects not appearing after re-login: Railway reissued GitHub App installations, and if the email in GitHub differs from the email in Railway, a re-login might be necessary.
We encountered this in practice: in one of our Spring Boot projects deployed on Railway, around 01:30 UTC on October 22, all requests to OpenRouter started failing:
401 Unauthorized on POST request for "https://openrouter.ai/api/v1/chat/completions": [no body]
org.springframework.web.client.HttpClientErrorException$Unauthorized
The balance was positive, and the code hadn't changed. Soon after, an email arrived from OpenRouter about the preemptive revocation of the key. The new key worked immediately locally, but it couldn't be deployed to production until Railway restored builds: the container continued to operate with the old, already revoked key. Later, the 401 changed to 402 — and it was still the old key, not a payment issue.
On October 23, a separate email arrived from Railway: our workspace was flagged as high-risk because variable names matched the pattern the attacker used to search for AI and cloud keys. Railway could not confirm whether our secrets were decrypted. By then, we had already rotated all keys, so no further action was needed — but this clearly shows that even variable names like OPENROUTER_API_KEY make a project a target.
If you also received a 401, it's not an issue with your code. Create a new key, update Variables, and redeploy the service. Railway has also published a guide on checking for unexpected OpenRouter usage.
🎯 Section 6. What Railway Users Should Do
Railway recommends everyone revoke and re-issue sensitive environment variables stored on the platform before October 21st, and to check the logs of relevant services. The key principle: creating a new secret isn't enough; the old one must be revoked on the provider's side, otherwise it remains valid.
1. Rotate All Secrets, Not Just AI Keys
Prioritize areas where you could lose money or data:
- Wallets and Payment Keys — If private keys for crypto wallets, seed phrases, or Stripe keys were stored in Railway, consider them compromised. Researcher Taylor Monahan, according to The Hacker News, reported fund drains from such wallets; Railway has not confirmed this. For crypto, key rotation means a new wallet and transferring funds. For Stripe, there's a Railway guide.
- Databases — Passwords for PostgreSQL, MySQL, MongoDB, or Redis. Step-by-step Railway guides: PostgreSQL, MySQL, MongoDB, Redis. If your DB is external (Neon, Supabase, etc.), change the password there.
- AI Providers — OpenRouter, OpenAI, Anthropic, xAI, etc., even keys that haven't been revoked yet. Guides for checking unexpected usage: OpenRouter, OpenAI, Anthropic, Grok.
- Cloud Credentials — AWS, GCP, S3, R2, Cloudinary. Guides: AWS, GCP.
- Email — SMTP, SendGrid, Resend.
- Application Secrets — JWT, session keys, OAuth client secrets, webhook secrets.
- Private Registries and GitHub — Docker Hub, GHCR, GitHub tokens. According to CoinInsider, GitHub tokens removed from Railway remain valid on GitHub until you revoke them yourself.
2. Restrict Public Access to Databases
Railway specifically advises removing public network access to databases. If the database is only accessible via the private Railway network, even a leaked password won't allow external connections. For most applications running in the same project, public access isn't necessary.
3. Check GitHub and Registries
- GitHub: Check your account or organization's audit log and the list of installed GitHub Apps — remove any you don't recognize. The Railway guide can help.
- Container Registry (Docker Hub, etc.): Check the audit log for unexpected pushes, pulls, or new tokens.
4. Replace Railway's Own Tokens
All Railway API tokens, as well as Project Tokens created before the incident, have been revoked. If you use the CLI, Public API, or `RAILWAY_TOKEN` in CI/CD (GitHub Actions, etc.), create new tokens and update them in your pipelines.
Railway has rotated its storage buckets itself: new credentials appear in the dashboard automatically, but services need a redeploy to pick them up. The procedure is in the guide on redeploying buckets and reconnecting GitHub.
5. Check Logs for Suspicious Activity
- AI Providers: Activity/Usage page — unfamiliar models, atypical volumes, requests at night.
- Payment Services: unexpected payments, new keys, or webhooks.
- Databases: connections from unfamiliar IPs.
- Cloud Storage and Accounts (AWS, GCP): unexpected uploads, deletions, or new resources.
To avoid unnecessary alarm: revoking bucket credentials on behalf of a Railway employee in audit logs is an expected action by the company itself, not unauthorized access to your account.
6. Deploy New Values
Generate new keys, save them in your password manager, enter them into Variables, and redeploy your services — new values are only applied after a deploy or restart. Builds for all Railway plans were restored on October 22nd.
7. Beware of Phishing
Following news like this, fake emails "from Railway" or "from OpenRouter" start appearing. Genuine emails are also sent: on October 23rd, Railway sent warnings to workspace owners at higher risk. However, the links in these emails go through tracking domains, so it's safer to verify information directly on Railway's blog and in your dashboard, rather than clicking links in the email.
Nuance: Database on the Same Railway Instance
Changing a database password updates the variables, but the application will only pick up the new password after a deploy. Therefore, the order is important: change the password and immediately deploy all services that use it — this minimizes downtime. If you also restrict public access to the database (step 2), the risk from the old password disappears even before rotation.
Nuance: Sealed Variables
Sealed variables cannot be read back through the UI, API, or CLI. However, neither the disclosure nor Railway's preliminary report clarifies whether they were affected by the incident — this is also noted in the review on daily.dev. Therefore, it's safer to rotate them as well.
How to organize secret management so that the next similar incident costs minimal time and money is covered in our guide "Secrets in SaaS: Storage, Rotation, and Incident Response Plan". A separate takeaway from this situation: do not store private wallet keys or keys without spending limits in the environment variables of any platform.
🎯 Section 7. What is Still Unknown
- Which specific variables were decrypted. Railway knows the upper bound — up to 1.85% of workspaces — but due to envelope encryption, they cannot identify the specific keys in affected projects. It's also unknown if sealed variables were affected.
- Technical details of the exploits. The general attack vector is known (build infrastructure → CI → environment variables), but Railway has not yet disclosed the specific vulnerabilities, citing ongoing investigations.
- Whether source code was accessed. Railway does not rule this out and promises to confirm in the final report.
- Whether crypto thefts were related to this incident. Railway has not confirmed Taylor Monahan's claim of ~$371,000 in thefts. A Symantec Threat Hunter Team analyst told The Hacker News that they see no evidence linking the observed activity to the incident.
- Who is behind the attack. Railway only states that the attacker had access to models with removed restrictions and significant computational resources, with the goal of model distillation. The case has been handed over to external forensics specialists and law enforcement.
Railway promises a final report after the investigation is complete. Updates are available on status.railway.com and in the Railway blog.
❓ Frequently Asked Questions (FAQ)
My service on Railway is working. Does this concern me?
Yes, if your secrets were stored in Railway environment variables before October 21st. The fact that your application is running doesn't mean your keys couldn't have been read. According to the preliminary report, environment variables were potentially compromised in at most 1.85% of workspaces, but Railway cannot determine which ones. Therefore, the company recommends rotating secrets for everyone, without waiting for a specific notification.
I received an email stating my workspace "is of interest to an attacker." Was I hacked?
Not necessarily. On October 23rd, Railway sent emails to owners of workspaces where environment variable names matched the pattern the attacker was using to search for AI and cloud keys. The company explicitly states they cannot confirm if your secrets were decrypted; this is a warning about increased risk, not confirmation of a leak. If you've already rotated your keys, no further action is needed. We received the same email and had already rotated everything by then.
Why is my application suddenly getting a 401 from OpenRouter?
Most likely, the key has been revoked. Railway asked providers to revoke potentially compromised keys, and even shared all keys from the platform with some of them. Therefore, an email about a key being disabled doesn't necessarily mean your specific key was leaked — but a disabled key cannot be restored, and a new one is required. According to OpenRouter's documentation, if compromise is suspected, the key should be deleted and a new one created. While you're at it, check your Activity for unauthorized usage (Railway's guide) and set a credit limit for your new key — this option is available.
Why did I get a 402 Payment Required error instead of a 401?
Based on our observations, if your balance is positive and the application continues to work with the old key, the 402 error is also related to a revoked key, not payment issues. We haven't seen an official explanation for this behavior, so test it this way: if a new key works locally, then the problem lies with the old key in the container, and it will be resolved after deploying with the new value.
I changed an environment variable in Railway, but nothing changed. Why?
Environment variables are only applied during deployment or container restart. Railway restored builds for all plans on October 22nd, so after changing the variable, simply redeploy your service.
Can I just put the new key directly in the code?
No. Changing the code still requires a deployment, and having a key in the repository is an additional risk. OpenRouter is a partner of GitHub secret scanning and will notify you about detected keys, but even after that, you must delete the key and create a new one: it will remain visible in the git history.
Which Railway tokens are no longer working?
Railway invalidated all API tokens, as well as Project Tokens created before the incident. If you use the Public API, CLI, or RAILWAY_TOKEN in your CI/CD, create new tokens and update them in your pipelines. The company also reissued storage bucket credentials: new ones appear in the dashboard, but the service requires a redeploy. Additionally, Railway reissued GitHub App installations — if your GitHub email differs from your Railway email, you might need to log in again.
Do I need to rotate sealed variables?
It's safer to rotate them. Sealed variables cannot be read back through the UI, API, or CLI, but neither the disclosure nor the preliminary report from Railway clarifies whether they were part of the incident's scope — daily.dev also points this out.
Why didn't Railway notify users immediately?
According to the company's explanation, providers were asked to revoke keys at 09:25 UTC on October 21st, before the public announcement. Railway deliberately waited to recommend rotating secrets until after they had evicted the attacker from their infrastructure: otherwise, new keys could have been compromised immediately after being entered.
Is it safe to use Railway now?
According to Railway, yes: the company has audited and redeployed all affected systems, transitioned to isolated single-tenant builders, and reissued internal keys twice. Final conclusions will be available in the final report. Regardless of the platform, it's advisable to restrict public access to your databases and avoid storing keys without spending limits in environment variables.
Is it true that cryptocurrency was stolen due to this incident?
There are no confirmations from Railway. Researcher Taylor Monahan claimed theft of approximately $371,000 from wallets whose private keys were stored on Railway; Railway has not confirmed this amount or its connection to the incident. Symantec Threat Hunter Team analyst told The Hacker News that they see no evidence linking the observed activity to this incident. If your wallet keys were stored on Railway, it's safest to assume they are compromised.
Sources: