Threat actors are actively exploiting CVE-2026-6875, a critical sandbox escape vulnerability in ServiceNow's AI Platform that enables unauthenticated code execution. Defused Cyber detected in-the-wild exploitation following public disclosure of the flaw.
The vulnerability carries a CVSS score of 9.5, reflecting its severity. It bypasses the sandbox environment designed to isolate AI Platform operations, allowing attackers to execute arbitrary code without authentication. This removes a fundamental security boundary protecting organizations using ServiceNow's AI tools.
ServiceNow has released patches to address the vulnerability. Organizations running the AI Platform should treat this as urgent. The combination of authentication bypass and code execution creates direct pathways for attackers to compromise systems, steal data, or deploy malware.
The threat landscape has shifted rapidly here. The gap between disclosure and active exploitation has compressed. Organizations that delay patching now face meaningful risk of compromise by sophisticated threat actors already scanning for vulnerable instances.
ServiceNow customers should verify patch deployment across all affected systems immediately. The lack of authentication requirements means attackers can target instances directly from the internet without needing valid credentials. Network segmentation provides limited protection given the remote nature of the attack vector.
This incident underscores a persistent pattern in enterprise platforms. Sandbox escape vulnerabilities represent high-value targets because they break out of intentional security constraints. When combined with unauthenticated access, the risk multiplies exponentially.
Organizations should also review access logs for evidence of exploitation attempts. Early detection of compromise matters significantly in limiting damage from arbitrary code execution.