Railway confirms breach of customer secrets, urges all users to rotate keys

What happened

Railway, the cloud platform developers use to build and deploy applications, has confirmed a cyberattack in which hackers accessed customer secrets. In a blog post published on Wednesday and reported by TechCrunch, the company urged every one of its users to revoke and rotate their API and cloud keys.

According to TechCrunch, Railway said the attackers may have reached encrypted environment variables on the platform and, in some cases, decrypted their contents. Environment variables typically store the passwords, tokens and API keys an application uses to talk to databases, cloud storage and other external services. Whoever holds those decrypted values can authenticate as the legitimate owner and walk into everything those services protect.

Limited breach, broad warning

Railway's initial analysis suggests the intrusion touched only a subset of users, but the company is still advising all users to rotate their credentials, a sign that it cannot yet rule out wider exposure. TechCrunch reports that Railway has not said how many users were affected, what the root cause was, or whether the attackers have made contact, for example with a ransom demand. The company did not immediately respond to a request for comment.

While the investigation runs, Railway has paused customer deployments, which effectively freezes the websites and apps hosted on its platform.

A wider campaign aimed at AI providers

Railway believes the breach was likely part of a broader campaign that targets platforms in order to compromise AI inference providers, TechCrunch reports. Attacks of this kind try to coax AI systems into leaking sensitive data through crafted prompts that slip past their guardrails.

The timing fits a pattern. Earlier in the week, OpenCode, an open source AI coding tool, disclosed that hackers stole API keys and other secrets from more than 1% of its customers. OpenCode attributed the incident to a technique it described as AI distillation, in which large numbers of automated agents probe a model to extract sensitive information. The U.S. cybersecurity agency CISA, meanwhile, has recently warned that China-based AI companies are running large-scale attacks against U.S. firms in order to improve their own models.

TechCrunch notes that the Railway attack appears unrelated to two critical PHP vulnerabilities disclosed this week under the name PHP2Shell, which pushed companies and governments to pull sites offline ahead of coordinated patch releases on Wednesday. Amazon security researchers said they found evidence that those PHP bugs were already being exploited four days before the patches shipped.

Why it matters

For a developer platform, few failures are as damaging as leaked environment variables. A single stolen key can cascade into database and cloud account compromises far beyond the original breach, which is why Railway is asking its entire user base, not just confirmed victims, to rotate credentials.

The decision to halt all deployments, trading availability across the whole customer base for containment, signals how seriously the company views the risk. The incident also fits an emerging pattern of attacks aimed at the AI supply chain, where prompts become the extraction tool rather than broken code. For engineering teams the practical steps are clear: rotate any keys stored on Railway immediately, and check whether those credentials are reused anywhere else.