Article 292: OpenCode Breach Exposes Stored AI API Keys for 1.4% of Console Users

The company behind OpenCode, the open-source coding agent, disclosed on 17 October what it called a large-scale distillation attack, traced to a legacy console system. The breach affected 1.4% of OpenCode Console users who had linked their own provider keys - OpenAI, Anthropic, Google and others. Some users report unauthorised charges on those accounts. OpenCode says it is working with the three providers and has told affected users to revoke and replace their keys.

Key Takeaway: A percentage this small will be filed as a minor incident, and that is the mistake. The asset taken was not data in the ordinary sense but spending capacity: a live key is a licence to run expensive models until someone notices the invoice. Worse, the stated motive - distillation - means the attackers were not after money but after outputs they could use to train a competitor model, which turns a key leak into an intellectual-property event that charges alone will not reveal. The practical lesson is boring and widely ignored: every developer linking a frontier key into a third-party gateway should have a hard spending cap set, because the cap is the only control that keeps working after the credential stops being secret.

Sources: Crypto Briefing, Cyera, 17 October 2026