The key facts
- Railway confirmed that hackers may have gained access to encrypted environment variables, and for some, to decrypted data
- The company urged all users to revoke and replace their keys; deployments have been paused
- Railway links the incident to a broader attack on AI providers; earlier, OpenCode suffered a similar breach
What happened
Railway, a platform that allows customers to build applications and services on its cloud development environment, confirmed a cyberattack in which hackers gained access to some customer secrets. In a blog post on Wednesday, the company said attackers 'may have gained access to encrypted environment variables and, for some of them, the decrypted contents.' Such variables typically contain passwords and API keys that developers' applications use to access other services — cloud storage and databases. With these secrets, an attacker can log into systems and access data as a legitimate owner.
The company's response
Railway said that, according to initial analysis, the incident affected only 'a subset of users,' but urged 'all users' to revoke and replace all of their keys so attackers cannot use them. During the investigation, the company paused customer deployments, effectively freezing their websites and applications. Railway did not disclose how many users were affected or the root cause of the incident; it did not respond to requests for comment, including about possible ransom demands.
Link to attacks on AI providers
It is not known how the attack was carried out or which Railway systems were compromised. The company said the breach is likely part of a broader campaign against platforms 'with the goal of compromising AI inference providers' — attacks in which AI systems are coerced through malicious prompts into revealing sensitive data, bypassing protective restrictions. The breach came shortly after the US agency CISA warned that China-based AI companies are conducting large-scale attacks on American companies to improve their own models. Earlier this week, the open-source AI coding tool OpenCode reported a similar breach: hackers stole more than 1% of customers' API keys and other secrets, and the attack was linked to 'AI distillation' — the mass extraction of information from AI models by agents.
Not related to PHP vulnerabilities
According to TechCrunch, the attack on Railway does not appear to be related to two critical vulnerabilities in the PHP language disclosed this week, which prompted companies and governments around the world to temporarily take websites offline to install fixes (PHP is used on roughly 70% of websites and public internet systems). Patches for the bug, dubbed 'PHP2Shell,' were publicly released on Wednesday following a globally coordinated campaign that allowed large organizations to protect their systems in advance. Amazon security researchers reported finding signs of exploitation four days before the patches were published, meaning the vulnerabilities are already being used in attacks.

