OpenCode disclosed a breach affecting 1.4% of Console users with external AI providers configured, exposing API keys for OpenAI, Anthropic, and Google in what it called a distillation attack.
Some users reported unauthorized charges exceeding $1,000. CellCog analyzed why real keys are targeted: they are verified, have payment history, and let fraudulent traffic blend as ordinary use, billed to the victim. Attackers continued attempts for at least 48 hours; frontier AI models refused to help analyze them when safety guardrails activated.