According to CoinMeta, following OpenCode and Railway, the programming tool Command Code announced that it had thwarted a large-scale distillation attack related to the Anthropic and OpenAI models. As a preventative measure, all API keys were revoked, and users were required to log in again to generate new keys. Over the past two weeks, OpenCode and Railway have also encountered similar attacks. On October 17th, the OpenCode cloud environment was compromised, affecting approximately 1.4% of users, and some third-party model keys may have been leaked, with users reporting losses ranging from several hundred to thousands of dollars. On October 21st, Railway was breached, posing a risk of environmental variable leaks in about 100,000 workspaces. Investigations revealed that hackers specifically targeted the API keys of the AI service, intending to use these keys to invoke models on a large scale. On October 14th, Zeabur was also invaded, with attackers searching for AI API keys and attempting to use the stolen AWS credentials to access Amazon Bedrock, but they were not successful.