Thought Leadership
Attackers breached OpenCode systems. Stored API keys for OpenAI, Anthropic, and Google may have been exposed.
Unauthorized access to a legacy system at developer platform OpenCode may have exposed sensitive API keys for leading AI providers. Affected users are already reporting unauthorized charges.
An attacker gained access to a legacy OpenCode system, potentially compromising credentials for third-party AI services stored in the cloud-based OpenCode Console. The console allows teams to store API keys centrally, enabling employees to use AI models without having direct access to the underlying credentials.
1.4 Percent of Console Users Affected
According to the company, the incident affects approximately 1.4 percent of Console users, specifically those with linked third-party providers. OpenCode has emailed affected customers, urging them to immediately revoke their keys through the respective providers’ developer dashboards and generate new ones.
According to co-founder Dax Raad, OpenCode’s desktop and command-line tools were not affected. These tools store API keys exclusively on the user’s device in a local SQLite database and do not transmit them to the company’s servers.
Suspected Model Distillation Campaign
Because API keys are bearer tokens, anyone who obtains them can run billable API requests at the victim’s expense. Deleting a key from the OpenCode Console is not sufficient, as this does not invalidate the key with the provider itself. Several developers have already reported unauthorized charges on social media, ranging from a few hundred to as much as USD 1,700.
OpenCode linked the security incident to a large-scale “distillation” campaign. In such campaigns, actors attempt to extract training data for competing AI models by submitting large volumes of queries to existing models. Raad said the service’s systems faced sustained attack attempts over a 48-hour period while the company implemented countermeasures and strengthened its security defenses.
(Editorial Team)
