Vulnerabilities
OpenAI Codex configuration flaw enables user-triggered RCE; patch now
Zero Day Initiative disclosed a Pwn2Own vulnerability in OpenAI Codex that lets an attacker run arbitrary code when a user opens a crafted page or file. The issue stems from insufficient sanitization of configuration settings and is tracked as CVE-2026-19590 with a CVSS 7.8 score. Defenders should apply OpenAI's update, restrict untrusted content, and monitor for unexpected local execution.
Zero Day Initiative published an advisory for a remote code execution flaw in OpenAI Codex, disclosed from Pwn2Own research. The vulnerability, assigned CVE-2026-19590, allows code execution in the context of the current user after a victim interacts with attacker-controlled content.
What happened
The flaw is in how Codex processes configuration files. Insufficient sanitization of configuration settings permits an attacker to influence system or configuration behavior. The attack requires the target to visit a malicious page or open a malicious file; no prior authentication is required.
Impact
- Affected product: OpenAI Codex.
- Severity: CVSS 7.8, vector AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H.
- Exploitation requires user interaction, but successful compromise can lead to full local code execution.
- OpenAI has released an update that addresses the issue.
Defenders should treat this as a high-priority desktop/client update, especially where developers or operators routinely open external repositories, prompts, or generated files.
What to do
- Apply OpenAI's Codex update immediately on all affected systems.
- Inventory developer workstations, CI agents, and lab machines that run Codex.
- Restrict opening untrusted pages, files, or configuration artifacts in Codex until patched.
- Review local execution logs, shell history, and process creation for unexpected commands after suspicious file or page access.
- Enforce least privilege for Codex users and isolate sensitive credentials from developer accounts.
Related vulnerabilities
OpenAI Codex RCE via Unsanitized Config Setting (CVE‑2026‑19592)
A remote code execution flaw in OpenAI Codex allows attackers to run arbitrary code when a user opens a malicious file or visits a crafted page. The vulnerability stems from missing sanitization of configuration settings and requires user interaction. OpenAI has released a patch; defenders should update immediately.
Read briefing
Kemp LoadMaster Uninitialized Memory RCE - CVE-2026-8037
Progress Software’s Kemp LoadMaster is vulnerable to a remote code execution flaw in the escape_quotes function. The issue requires authentication and allows attackers to run code as root. Apply the vendor patch immediately and verify configuration hardening.
Read briefing
CVE-2026-69918: Windows Core Messaging Heap Overflow Enables Local Privilege Escalation
A heap‑based buffer overflow in the Windows Core Messaging component can be triggered by an authenticated user to gain elevated privileges on the local system. The flaw is exploitable only with authorized access and is addressed in the latest cumulative update. Defenders should apply the patch immediately and enforce least‑privilege policies.
Read briefing
CVE‑2026‑83758: Windows Error Reporting Heap Overflow Enables Local Code Execution
A heap‑based buffer overflow in Windows Error Reporting can be triggered by an authorized user to run arbitrary code on the affected system. Microsoft has issued a patch; administrators should apply it or disable the service to mitigate.
Read briefing