OpenCode says breach affected 1.4% of Console users with linked AI providers

OpenCode said on Oct. 18, 2026, that a security breach linked to what it described as a “large scale distillation attack” affected 1.4% of OpenCode Console users who had connected external AI providers. The company named Anthropic, OpenAI and Google, and said it was working with those providers to respond.

The disclosure covers only Console users with linked providers, not everyone who has installed OpenCode’s command-line coding agent. OpenCode said it emailed every affected user that morning and that users who did not receive a message were not impacted. Those are the company’s statements; it has not published a root cause, identified the data involved or said what the attackers accessed. The incident was still ongoing when StepSecurity published its account.

The unanswered questions matter because linked provider accounts let OpenCode call models on a developer’s behalf. StepSecurity, a software-security company, said exposed credentials could allow unauthorized inference and lead to unexpected provider charges or access to data reachable through those credentials. It did not report that any of those outcomes had occurred in this breach.

OpenCode has not listed affected software versions, and StepSecurity cautions that the installed CLI version alone does not establish whether someone was affected. The relevant factors, according to its analysis of the vendor’s statement, are whether a developer had an OpenCode Console account and had linked an external provider. StepSecurity says provider credentials are stored locally in an OpenCode authentication file, with different file paths for macOS and Linux versus Windows.

For organizations trying to establish whether employees use the tool, StepSecurity recommends checking developer machines for OpenCode and then determining whether providers were linked. Its free, open-source Dev Machine Guard can scan machines for coding agents and other developer tools; StepSecurity says the software detects OpenCode and reports its version and location. The company also offers enterprise customers an inventory view through its console, but that inventory does not reveal whether a user has a Console account or which providers are connected.

StepSecurity recommends that developers check for an email from OpenCode, identify linked providers and rotate the associated API keys, then review provider usage for unfamiliar requests. Those are precautionary steps from the security firm, not a detailed remediation plan issued by OpenCode. The company’s public account had not provided further incident details.