OpenCode: Console breach exposed stored AI provider API keys, ~1.4% of users hit
Severity: high · Patch status: no-fix-needed · Since 2026-10-17
OpenCode, the open-source AI coding agent, disclosed on 2026-10-17 that attackers broke into an older system behind its hosted Console and may have read the Anthropic, OpenAI and Google API keys stored in affected workspaces; the earliest possible key exposure was 2026-10-17 00:30 UTC. The company says about 1.4% of Console users were affected, that it emailed them directly (no email means not affected), and that the open-source app and CLI were not touched. It framed the event as a large-scale distillation attack, but it has not published how the intrusion worked, how keys were stored or the full extent of misuse. Victims reported unauthorised charges from several hundred to roughly 1,700 USD overnight, and some Google projects were suspended for abuse. Anthropic and Google blocked the keys they believed were exposed, OpenAI refunded API customers for credits consumed, and OpenCode says refunds are under way while the incident remains ongoing. Wiz counts it among at least nine AI developer platforms breached in October 2026 (Black October). Affected users should revoke and replace linked provider keys at each provider and review usage and billing.
What to do
- If you received an OpenCode breach email, treat every provider key linked in Console as compromised; no email means OpenCode says you were not affected
- Revoke and replace linked Anthropic, OpenAI and Google API keys directly at each provider
- Review provider usage and billing dashboards from 2026-10-16 onward for unexpected inference or spend, and ask OpenCode and providers about refunds
- Check Google Cloud for projects suspended for abusive activity and for new API keys or service accounts you did not create
- Set spend limits and alerts on provider accounts, and avoid storing shared team keys in a hosted console where possible
- Maintain an inventory of AI coding agents and the credentials linked to them
Sources
All live incidents on Threat Watch
