OpenCode confirms a breach tied to a 'large-scale distillation attack' hit 1.4% of Console users with linked OpenAI/Anthropic/Google keys; working with the three labs to remediate

Why it matters

【Key points】

OpenCode, the open-source AI coding agent, confirmed on its official X account on Oct 17 (US time) a security breach the previous night "related to a large scale distillation attack" that affected 1.4% of OpenCode Console users — a subset who had linked external model providers (OpenAI, Anthropic, Google). It says all affected users were emailed that morning, anyone without an email was not impacted, the incident is ongoing, and it is working with Anthropic, OpenAI and Google to remediate.

Hours earlier, users posted OpenCode notices describing unauthorized access to a legacy system being decommissioned and possible exposure of stored third-party API keys; several reported unexpected charges of tens to hundreds of dollars and one a suspended Google project.

【Verifiable】

OpenCode statement: https://x.com/opencode/status/2111519637049319818

First version: https://x.com/opencode/status/2111516673953935833

Matt Johansen's early roundup: https://x.com/mattjay/status/2111476982580211851

Status page: https://status.opencode.ai

【Our read】

1. So far there is only one X statement — no blog post, advisory or media coverage; attack path, scope and timeline are undisclosed.

2. "Distillation attack" suggests stolen keys were used to mass-query frontier models for training data, billed to users' own accounts.

3. Anyone who stored provider keys in OpenCode Console should revoke and rotate them, set spend caps and check usage, email or not.