NeuralStackAn intruder reached a legacy system at OpenCode on the night of October 16, potentially exposing third-party AI provider keys saved in the company’s hosted Console service. The affected credentials belonged to OpenAI, Anthropic, and Google accounts linked by users to their workspaces.[2]
OpenCode Console allows a workspace to store an outside provider key so the service can make model calls on behalf of team members without revealing the key to them. The company said the incident affected 1.4% of Console users, all of whom had an external provider linked.[2]
OpenCode notified the affected users by email on the morning of October 17 with the subject line “Action Required: Potential provider key breach.” The company instructed them to revoke the exposed keys with each provider and replace them. Users who did not receive the email were not affected, OpenCode later said.[2]
The exposure concerns keys pasted into Console rather than credentials stored by OpenCode’s command-line and desktop applications. Those applications keep provider keys on the user’s machine in a SQLite database, according to the supplied report.
The incident arrives as attackers increasingly combine access to software infrastructure with AI services and automation. A separate campaign attributed to the financially motivated initial access broker CyberXero used Anthropic’s Claude Code alongside Cobalt Strike while targeting WordPress sites, e-commerce platforms, and Ukrainian critical infrastructure.[4]
Why this matters
AI provider keys can grant access to model services under an organization’s account, making them valuable operational credentials rather than ordinary configuration data. A breach involving a hosted workspace therefore creates potential exposure beyond the workspace itself, particularly when multiple team members rely on centrally stored keys.
The broader security picture is also becoming more difficult to contain. Researchers reported the PEERSTITCH malware being delivered through exploitation of a FortiOS flaw, while CISA urged organizations to update Ivanti Connect Secure after exploitation of an authentication-bypass vulnerability.[1]
OpenCode’s immediate advice is straightforward: affected users should revoke the old provider credentials and issue replacements. The episode is likely to intensify scrutiny of how AI platforms store, isolate, monitor, and rotate customer-supplied keys as organizations expand their use of hosted model tooling.