OpenCode Console Breach May Have Exposed OpenAI, Anthropic, and Google API Keys

OpenCode says an intruder reached a legacy system on the night of October 16 and may have exposed the third-party AI provider keys that users had saved in OpenCode Console.

Console is OpenCode's optional hosted service: a workspace stores its own OpenAI, Anthropic, or Google key there, and Console uses that key to call the provider so team members can work with the models without ever seeing the key itself.

The affected group is 1.4% of Console users, all of whom had an outside provider linked. On the morning of October 17 OpenCode emailed them under the subject line "Action Required: Potential provider key breach," urging them to revoke those keys with each provider and replace them. Anyone who got no email, the company said, was not affected.

Only Console keys are at risk

The exposure concerns keys pasted into Console. Keys added through OpenCode's command-line or desktop apps stay on the machine in a local SQLite database. As co-founder Dax Raad wrote in a reply on X, the incident involves "the OpenCode Console, not OpenCode itself," and those keys "never come to us."

A provider key is a bearer credential: whoever holds a copy can run billable requests against the owner's account until the provider revokes it. Deleting the key from Console does not revoke it, and a Console spend limit won't help either, since Console only sees requests that pass through its own gateway.

What to do

OpenCode and the providers advise revoking the key in the provider's dashboard and creating a new one, updating every application and script that used it, reviewing billing for requests you did not make going back at least to October 16, and setting a spend limit while giving each tool its own key.

Some users say they have already paid for usage they never ran. Dealroom reports that Colin Parsons wrote on X his OpenAI key was used for $281 and his Anthropic key for $168, while Preston Zeller put his losses at $1,700. Notebookcheck cites Reddit users claiming $400 of unauthorized Anthropic usage and about €200 of Google usage, with the project then suspended.

OpenCode's own wording is more cautious: the keys "may have been exposed," and it has not said any key was copied or used.

A "large-scale distillation attack"

In its October 17 post OpenCode called the incident ongoing, said it was working with Anthropic, OpenAI and Google, and described the breach as "related to a large-scale distillation attack." It gave no further detail — no names, no description of the legacy system, no count of keys affected.

Distillation is the term providers use for campaigns that pull a model's answers en masse to train a rival. In September Anthropic noted such campaigns reach its models through networks of fake accounts built on stolen cards, credentials and harvested API keys. On October 18 Raad wrote that attempts to breach OpenCode's systems had been "non stop" for the past 48 hours, with "a side effect" that "everything is getting hardened."

Source: The Hacker News