OpenCode Console breach exposes third-party AI API keys

OpenCode has confirmed a security breach involving its Console service, which has impacted approximately 1.4% of its users. The incident occurred when unauthorized actors accessed a legacy system that was in the process of being decommissioned.

The breach specifically targeted third-party API keys stored within user workspaces, including those for OpenAI, Anthropic, and Google. Attackers reportedly used these keys in a “large scale distillation attack,” a method where frontier model outputs are harvested to train competing models. This activity has resulted in unauthorized usage charges for some victims, with reports of expenses exceeding $1,000.

OpenCode co-founder Dax Raad clarified that the compromise was limited to the OpenCode Console and did not affect the open-source application itself or the command-line tools. The company has notified affected users via email, advising them to immediately revoke and replace their provider keys. OpenCode is currently working with providers like OpenAI to address the situation and facilitate the reimbursement of unauthorized costs incurred by users.