A cybersecurity breach involving cloud hosting provider Railway has exposed a growing financial risk in the artificial intelligence industry. Hackers stole customer credentials and used them to access AI services, generating an estimated $740,000 in unauthorized charges before the company contained the attack.
According to Railway's incident report published November 10, the attack began on October 20, when hackers exploited a previously unknown vulnerability in its build infrastructure. The attackers used the access to move through internal systems, examine stored customer credentials, and target API keys belonging to AI providers.
Railway said up to 1.85% of its customer workspaces, approximately 100,000, could have had environment variables decrypted. This is the maximum potentially affected number, not a confirmed count of customers whose credentials were stolen.
How Hackers Turned Stolen AI Keys Into Money
The attackers specifically searched for AI provider credentials, including API keys used to access services from companies such as OpenAI and Anthropic. These keys allow applications to connect to AI models and pay for computing resources through usage-based billing.
Once the attackers obtained access, they could make requests to AI services using compromised customer accounts. Instead of paying for their own computing resources, they effectively transferred the costs to legitimate account holders.
This type of attack is sometimes called LLM jacking, where cybercriminals steal access to large language models and use the victims' accounts to run AI workloads or resell computing access.
The Growing Security Problem Behind AI
The Railway incident highlights a security weakness that is becoming increasingly important as businesses integrate AI into their operations. Companies frequently store API keys, passwords, and other sensitive credentials in cloud platforms to connect applications and services.
If attackers gain access to those credentials, they can potentially use expensive AI models, cloud infrastructure, and other paid services without authorization. The resulting damage is not limited to stolen information; businesses can also face unexpected bills, disrupted operations, and costly security investigations.
Railway said it has closed the vulnerabilities used in the attack, revoked affected credentials, and introduced additional infrastructure protections. The company is also coordinating with AI providers to recover unauthorized charges and has committed to covering costs that providers refuse to refund.
What Businesses Should Watch
As AI adoption expands, protecting access credentials is becoming just as important as protecting customer databases and financial information. Businesses should regularly rotate API keys, restrict access permissions, monitor unusual usage, and establish spending limits and billing alerts wherever possible.
The larger concern is that stolen AI credentials can quickly become a source of computing power for cybercriminals. With AI services increasingly priced according to usage, a compromised account can create substantial financial losses before the victim even realizes something is wrong.
The Railway breach is a reminder that the AI industry's rapid expansion is creating new opportunities for cybercrime. As more companies depend on AI-powered services, securing the credentials behind those systems will become a critical part of protecting both their operations and finances.
Discussion
Thoughts from readers and local market watchers.