CVE-2026-14898 describes an information exposure vulnerability in the OpenAI Codex desktop app for macOS. The app rendered remote images from Markdown in model responses. An attacker who could place an indirect prompt injection in content processed by Codex—such as results from connected tools or other untrusted sources—could induce the model to construct a remote image URL containing sensitive information from the Codex session. The app automatically fetched the remote image URL when rendering the response, thereby sending the embedded sensitive data to an attacker-controlled server without requiring user interaction. The vulnerability could lead to exfiltration of secrets including API keys, source code, and connected tool data. No integrity or availability impacts were demonstrated, and there is no evidence of exploitation in the wild. No patch or official remediation has been confirmed as of the published date.

Successful exploitation could result in the unauthorized disclosure of sensitive information accessible within the Codex session, including API keys, source code, and data returned by connected tools. There is no impact on system integrity or availability. No known exploitation in the wild has been reported.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is available, users should exercise caution when processing untrusted content in the Codex desktop app for macOS. No vendor advisory or patch links are currently provided.