OpenCode Breach: Rotate Your OpenAI, Anthropic and Google API Keys Now

What happened

Anomaly Innovations, the company behind the open-source coding tool OpenCode, began emailing users late on October 17 that provider API keys stored in OpenCode workspaces may have been exposed through a legacy console system the company was in the process of decommissioning. The earliest exposure the company could identify was roughly 22 hours before that first notice went out.

The notice covers keys for OpenAI, Anthropic, and Google — the same keys developers use to pay for model calls inside their coding workflows. Two users have already reported unauthorized Claude API usage, and about $100 in fraudulent charges has been tied to the incident so far.

Notably, the detailed timeline of the exposure came from a developer in the community rather than from the company itself — a transparency gap that has not gone unnoticed.

Why this matters

API keys are the financial plumbing of the AI economy. A leaked key is not just a security event; it is a direct line to your billing account. Attackers routinely scan for exposed AI keys because a single key can be converted into thousands of dollars of inference spend in hours, and the charges land on the key owner's card.

This incident also highlights a familiar failure pattern: the vulnerable system was a legacy console being retired — exactly the kind of half-forgotten infrastructure where credentials pile up and monitoring is thinnest.

What this means for your business

1. Audit where your API keys live. If anyone on your team used OpenCode with production provider keys, those keys must be revoked and reissued immediately — directly inside each provider's dashboard, not through links in the notification email.

2. Check billing and usage logs back to October 17. Look for usage spikes, unfamiliar projects, or calls outside your team's working hours. Two users caught the fraud early only because they were watching.

3. Set spending limits and usage alerts now. Every major AI provider lets you cap spend per key. A leaked key with no cap is an unlimited-draw account for whoever finds it.

4. Separate dev and production keys. Keys used by coding agents and experimental tools should never share a billing account with production systems. Scope keys narrowly and rotate them on a schedule — not only after a breach.

5. Don't trust retired systems to stay dead. If your company is decommissioning any system that ever touched credentials, treat the wind-down window as a live risk: audit stored secrets first, revoke second, shut down third.