OPENCODE has disclosed a security incident described as a large-scale distillation attack that affected a subset of its Console users—specifically 1.4 per cent of those with linked AI providers. The vendor said the incident is ongoing and that affected users were notified by email, naming Anthropic, OpenAI and Google as the providers involved. At the time, details on what data was accessed, whether API keys were exposed, or what exactly “distillation attack” entails were not published.
The post notes that the impact is limited to Console users with linked provider accounts, not all OpenCode users, and no root cause has been shared publicly yet.
What this means in practice is that if an attacker obtained credentials tied to linked AI providers, they could potentially run inferences against those provider accounts, which could manifest as unexpected spending or access to data reachable by the keys. The article stresses that many security teams lack visibility into which developers have OpenCode active and which providers are linked, creating a preparatory gap.
Steps outlined to respond focus on identifying affected developers, verifying whether OpenCode was installed and which providers are linked, and rotating the associated API keys. The guidance includes commands for locating the local auth file (for example, cat ~/.local/share/opencode/auth[.]json on Linux/macOS, or the Windows path), and emphasises reviewing provider usage dashboards for anomalous activity. The article also offers methods for non-StepSecurity customers to discover installed OpenCode via a one-off scan and a 14‑day free trial.