A security incident involving OpenCode's hosted Console platform may have exposed API credentials belonging to users who connected third-party AI services such as OpenAI, Anthropic, and Google, prompting the company to urge affected customers to immediately rotate their keys.
According to OpenCode, an attacker gained access to a legacy internal system during the night of October 16, 2026. While the company has not confirmed that any credentials were stolen or abused, it warned that API keys stored within Console workspaces may have been exposed and should be considered at risk until proven otherwise.
OpenCode Console serves as an optional cloud-hosted collaboration layer that allows teams to use AI models through shared workspaces. Instead of distributing credentials among employees, organizations can store provider keys within Console, which then handles requests to OpenAI, Anthropic, or Google on behalf of workspace members. This design prevents users from directly viewing or accessing the underlying API credentials.
The company stated that approximately 1.4% of Console customers were potentially impacted, specifically those who had linked external AI-provider accounts. On the morning of October 17, OpenCode sent notifications titled “Action Required: Potential Provider Key Breach” to all affected users, instructing them to revoke and replace credentials associated with their AI providers. Users who did not receive the notification were reportedly unaffected.
OpenCode emphasized that the incident was limited to its hosted Console environment and did not impact credentials stored locally through its standalone applications. API keys configured directly within OpenCode’s desktop or command-line clients remain stored on the user's device in a local SQLite database and are not transmitted to OpenCode’s infrastructure.
The distinction is significant because API keys function as bearer tokens. Anyone possessing a valid key can submit requests and generate charges against the associated account until the credential is revoked. Simply deleting the key from OpenCode Console does not invalidate it, and Console-level spending controls cannot prevent unauthorized usage occurring outside of OpenCode's own infrastructure.
As a precaution, affected organizations have been advised to:
- Immediately revoke exposed API keys through the provider’s dashboard.
- Generate replacement credentials and deploy them across applications, automations, and development environments.
- Review account activity and billing records for suspicious usage dating back to at least October 16.
- Configure spending limits, alerts, and separate keys for individual applications to reduce exposure and improve auditing.
Several users have already reported unexpected charges that may be linked to the incident. Public posts cited by industry reports describe alleged unauthorized usage ranging from hundreds to thousands of dollars across OpenAI, Anthropic, and Google accounts. However, these claims remain user-reported accounts, and OpenCode has not verified that the charges resulted directly from the breach.
The company’s official position remains that the affected credentials may have been exposed, rather than confirming that any keys were copied, distributed, or abused by attackers.
In a statement published on October 17, OpenCode described the intrusion as being connected to what it called a large-scale AI model distillation campaign. Model distillation refers to efforts by threat actors or competing organizations to collect massive volumes of responses from advanced AI systems in order to train alternative models. Access to stolen API keys can significantly reduce the cost of such operations while obscuring the attackers’ identities behind legitimate customer accounts.
Despite referencing the broader distillation activity, OpenCode has not disclosed details regarding the attackers, the compromised legacy system, the initial access method, the number of credentials stored on the affected platform, or the total customer count represented by the reported 1.4% exposure rate.
The situation remained active as of October 18, with OpenCode co-founder Dax Raad stating that attackers had continued probing the company's infrastructure for nearly two days following the breach. The company said it was working alongside OpenAI, Anthropic, and Google while simultaneously hardening its systems and implementing additional security controls in response to the ongoing attack activity.
The incident highlights the growing risks associated with centralized storage of AI-provider credentials, particularly as API keys have become valuable targets for cybercriminals seeking to monetize unauthorized AI usage, conduct large-scale automated operations, or support model-training campaigns using stolen resources.
Found this article interesting? Follow us on X(Twitter) ,Threads and FaceBook to read more exclusive content we post.
