OpenCode has warned that an intruder may have accessed third-party AI provider keys stored in its Console service after reaching a legacy system on October 16. The company said the incident affected 1.4% of Console users who had linked external AI providers to their workspaces.
OpenCode emailed affected users on October 17 with the subject line ‘Action Required: Potential provider key breach.’ The company asked them to revoke their existing keys through their providers and generate replacements. OpenCode said users who did not receive the email were not affected.
The exposed credentials could include keys for OpenAI, Anthropic and Google. Console stores these keys to let teams access AI models without sharing the underlying credentials with individual members.
Also Read: ASOS Cyber Breach: Hackers Accessed Customer Data Through Employee Account
OpenCode co-founder Dax Raad clarified that the incident involved the hosted Console service rather than the company's command-line and desktop applications. Those applications store provider keys locally in a SQLite database.
API keys act as credentials for accessing AI services. Anyone who obtains a valid key could make requests that generate charges on its owner's account. Removing a key from Console does not invalidate it, so users must revoke it through the provider's own dashboard.
OpenCode has advised affected users to replace compromised or potentially exposed keys, update applications that use them and review provider billing records for suspicious activity dating back to October 16. The company has not confirmed when the unauthorized access began or whether attackers copied or used any keys.
Some users have reported unexpected charges. Reports cited losses of USD 281 on an OpenAI account and USD 168 on an Anthropic account, while another user reported USD 1,700 in losses. These figures are individual claims, not confirmed totals from OpenCode.
The company described the incident as ‘related to a large-scale distillation attack’ and said it was working with the three AI providers. It has not explained the connection between the attack and the exposed credentials.