If you connected API keys from Anthropic, OpenAI, or Google to the cloud OpenCode Console, check your billing balance immediately. The project team has confirmed a data breach: tokens belonging to 1.4% of OpenCode Console users were compromised.
Attackers targeted an outdated legacy server for the console that was already in the process of being decommissioned. Their goal was a model distillation attack: after harvesting developer keys, hackers spun up synthetic training-dataset generation on someone else's dime. Engineers reported overnight charges ranging from $168 to $1,700.
The vulnerability affected only the cloud console. Developers running the local OpenCode CLI with tokens stored in environment variables on their own machines are safe.
If you used the web console, revoke those keys in your LLM providers' dashboards and reissue new ones with strict spending caps. Earlier, we covered how distillation attacks work and the risks of open AI tools, and this incident is another stark reminder: never hand long-lived master keys over to third-party web services.
![Node.JS [ru]](/api/digests/it_development/daily/20261021/assets/sources/we-use-js.jpg?v=449b8486fe24bd19bc7e6937ea5a143242b0d8bdef73ba6df994edf43501a455)