OpenCode Console breach: what to do if your OpenAI, Anthropic or Google key was exposed

Published October 25, 2026. Facts below are only what TechCrunch, The Hacker News, BleepingComputer, and eSecurity Planet reported from OpenCode and the providers.

What was reported

The Hacker News reports that an intruder reached a legacy system at OpenCode on the night of October 16 and may have exposed third-party provider keys saved in OpenCode Console. [2] eSecurity Planet describes the same notice as unauthorized access to a legacy system OpenCode was decommissioning, with third-party API keys stored in customer workspaces. [4]

OpenCode said the breach affected 1.4% of Console users, a subset who had an external provider linked, and that it emailed all of them. Users who did not get the email were not impacted, OpenCode said. [1] [2] [3]

OpenCode described the breach as related to a large-scale distillation attack. It has not said how the intrusion and that attack connect, and TechCrunch reports that the company did not specify how the distillation happened. [1] [2] OpenCode's own wording, as reported by The Hacker News, is that the keys may have been exposed. It has not said that any key was copied or used. [2]

On October 19, TechCrunch reported that OpenCode said the incident remained ongoing. [1]

1. Were you affected?

Search mail from the morning of October 17 for the subject Action Required: Potential provider key breach. The Hacker News reports that subject line, and that anyone who did not get the email was not affected. [2]

eSecurity Planet reports that a follow-up notice named the stored credentials by workspace, provider, and their final four characters. [4] BleepingComputer and eSecurity Planet both report that the notice set the earliest possible exposure of those keys at October 17, 2026, 00:30 UTC. eSecurity Planet also reports that the notice said no other data appeared to have been compromised. [3] [4]

Keys used only in the local command-line or desktop app were not part of this breach. Those apps keep provider keys on the machine. Raad wrote that this was the OpenCode Console, not OpenCode itself, and that those keys never come to OpenCode. [1] [2]

2. Revoke and rotate at each provider

Revoke the key in the provider's own dashboard and create a new one. Then update every application and script that used the old key. [2]

Deleting the key in Console does not revoke it. A Console budget cannot cap use outside Console's gateway, because Console only sees requests that pass through that gateway. [2]

• OpenAI, Anthropic, and OpenRouter. The revoke, rotate, usage check, and limit steps are in the first 10 minutes guide. OpenRouter was not named in these OpenCode reports. The same revoke rule applies if you reused one of these keys there.
• Google. TechCrunch reports that Google blocked the API keys it believed were exposed. eSecurity Planet reports that Google blocked exposed Gemini credentials and that users were advised to generate a replacement in AI Studio and examine account activity. [1] [4]

TechCrunch reports that Anthropic blocked the API keys it believed were exposed. eSecurity Planet reports that Anthropic revoked Claude API keys believed to be affected. A block from the provider does not finish the job: create a new key for anything that still needs access, and retire the old one yourself. [1] [4]

3. Check per-key usage from October 16 on

Review usage and billing at the provider for requests you did not make, going back at least to October 16. The Hacker News reports that OpenCode has not said when the unauthorized access began. [2] The notice's earliest possible exposure, October 17 at 00:30 UTC, falls inside that window. [3] [4]

Use the workspace, provider, and last four characters from the notice to find the key. Keep the export or the screenshots. This page does not prescribe a console click-path. The providers' own usage pages are the record.

4. Ask for a refund or credit

OpenAI said it worked with OpenCode to restore access and refund its API customers for credits used in the attack. [1] eSecurity Planet reports that on October 19 OpenAI said affected API customers would receive refunds for credits consumed during the attack. [4] OpenCode said refunds for unauthorized charges were underway. [1]

When you write, send the key name or the last four characters, the time you revoked it, and the usage you believe was not yours. Do not send the secret.

These sources do not describe a matching refund from Anthropic or Google. Contact that provider with the same usage details if the charges are there. Treat anything short of OpenAI's stated credit refund as a request.

5. Prevent the next one

• One key per tool. A key pasted into Console then maps to one provider key, and the notice's last four characters tell you which one to revoke. [2]
• A daily cap per key. A budget inside Console cannot cap requests that skip its gateway. [2] The daily budget guide covers what OpenAI, Anthropic, and OpenRouter can set, and what they cannot. It does not cover Google.
• Alerts and a kill switch. BillCurfew emails on a spike or a daily cap and can kill the key at OpenAI, Anthropic, or OpenRouter. Manual kill is on every plan, including Free. Google and Gemini are outside that list.

FAQ

Sources

1. TechCrunch, Hackers hit AI coding tool OpenCode, running up charges on users' AI accounts. https://techcrunch.com/2026/10/19/hackers-hit-ai-coding-tool-opencode-running-up-charges-on-users-ai-accounts/
2. The Hacker News, OpenCode Console Breach May Have Exposed OpenAI, Anthropic, and Google API Keys. https://thehackernews.com/2026/10/opencode-console-breach-may-have.html
3. BleepingComputer, OpenCode hack exposes API keys amid AI distillation attacks. https://www.bleepingcomputer.com/news/security/opencode-hack-exposes-api-keys-amid-ai-distillation-attacks/
4. eSecurity Planet, OpenCode Breach Exposes AI API Keys. https://www.esecurityplanet.com/artificial-intelligence/news-opencode-breach-ai-api-keys/