OpenCode Console suffers security breach in AI distillation attack

A security breach affected 1.4% of OpenCode Console users, according to a statement released by the company on Saturday, October 17. The incident, which occurred the previous evening, was attributed to a large-scale artificial intelligence distillation attack—a technique that uses responses from one model to train another.

The affected group consisted of users who had external providers linked to the service. The company sent email alerts that same morning, stating that anyone who did not receive the message was not affected.

OpenCode reported that the issue was still ongoing and that it was collaborating with Anthropic, OpenAI, and Google to contain it. The company did not identify the perpetrators or provide details on how access was obtained.

A security breach occurred last night related to a large-scale distillation attack that affected 1.4% of OpenCode Console users.

We sent email alerts to everyone this morning. If you did not receive an email, you were not affected—it was a subgroup of users with external providers linked.

The incident is still ongoing, so we cannot share more details at this time, but we are working with Anthropic, OpenAI, and Google to resolve the issue.

— OpenCode (@opencode) October 17, 2026

What is an AI distillation attack?

AI distillation involves using the responses generated by one model to train another, transferring some of its capabilities. When done with permission, it can be used to create smaller, more cost-effective systems. The problem arises when data collection is done without authorization or violates the provider’s terms.

OpenCode itself classified the incident as a distillation attack. The company did not explain how the attempt to obtain responses was related to the security breach, nor did it provide evidence that any model was actually trained with the material obtained.

The role of OpenCode Console

OpenCode is an open-source, AI-based programming agent available via terminal, desktop application, and extension for development environments. It enables interaction with models from different providers.

The Console serves as a gateway to models selected by the team, with usage-based billing. Its adoption is optional, and the documentation allows users to connect the agent to other providers. Therefore, the percentage disclosed refers only to Console users, not all OpenCode users.

Information not yet disclosed

The statement does not provide the total number of affected accounts, what data was accessed, or whether there was exposure of code, conversations, or credentials. There is also no estimate of financial loss or a deadline for completing the response to the incident.

The involvement of Anthropic, OpenAI, and Google in the response does not imply that these companies’ systems were compromised. The collaboration is limited to resolving the problem reported by OpenCode.

Frequently Asked Questions

Were all OpenCode users affected?

No. The incident affected 1.4% of Console users, specifically those with external providers linked.

How do I know if my account was affected?

OpenCode sent email alerts on the morning of October 17. Anyone who did not receive the message was not affected.

What does it mean to distill an AI model?

It is the process of using one model’s responses to train another, reproducing some of its capabilities. The technique can be legitimate, but it becomes problematic when it involves improper access or collection.

Were Anthropic, OpenAI, and Google also hacked?

The statement does not say this. OpenCode said it is working with the three companies to resolve the incident.

Is there confirmation of code or conversation leaks?

The statement does not specify what information was accessed or confirm exposure of code or conversations.