Railway breach exposed secrets in up to 100,000 workspaces
An attacker spent nearly two days inside the hosting platform hunting for AI provider API keys to use in model distillation, decrypting environment variables along the way.
Railway published a preliminary report on 22 October saying an attacker breached internal systems and decrypted some customers’ environment variables. Monitoring flagged abnormal activity at 13:58 UTC on 20 October, the attacker escalated privileges and reached CI credentials over the following hours, and Railway confirmed it had shut the intruder out by 03:32 UTC on 22 October.
Railway says at most 1.85% of all workspaces, around 100,000, had environment variables potentially compromised. The attacker was specifically after AI provider API keys to run distillation attacks, copying a larger model’s behaviour by querying it heavily. GitHub, GitLab, Anthropic, OpenAI, OpenRouter, Stripe and Google were all notified so they could revoke affected keys. Railway is telling customers to rotate anything stored on the platform before 21 October, remove public network access from databases, and check GitHub and container registry audit logs for anything unusual.
Why it matters: if you store secrets in a platform’s dashboard rather than your own vault, they’re only as safe as that platform’s internal access controls. Rotate your Railway credentials now rather than waiting for a direct notification, and check what else a breach there could have reached, including which third-party services you’d linked by API key.
The caveat: Railway says only customers with confirmed impact get a direct email; everyone else is relying on this public report to know whether to act.