The enterprise software landscape has been sent into a state of high alert following the disclosure and rapid weaponization of a critical remote code execution (RCE) vulnerability, tracked as CVE-2026-6875, affecting the ServiceNow AI Platform. With ServiceNow powering the operational backbones of 85 percent of Fortune 500 companies and facilitating over 100 billion workflows annually, the discovery of a pre-authentication vulnerability has sent shockwaves through the cybersecurity community.

The flaw allows unauthenticated attackers to gain complete control over self-hosted ServiceNow instances, effectively bypassing security barriers that were intended to isolate user-supplied data from the underlying system architecture. As of mid-July 2026, threat intelligence agencies have confirmed that the vulnerability is being actively exploited in the wild, leaving administrators of self-hosted environments in a race against time to apply necessary patches.

The Genesis of the Discovery: A Technical Deep Dive

The vulnerability was identified by researchers at Searchlight Cyber, who revisited the ServiceNow codebase nearly two years after their initial discovery of a similar critical flaw. Their latest report highlights an sophisticated "sandbox escape" technique that leverages the platform’s internal GlideRecord query API.

The Mechanism of Failure

ServiceNow’s GlideRecord API is a ubiquitous component of the platform, utilized across hundreds of functions to handle data queries. The system design includes a feature that permits developers to prefix query arguments with javascript:. When this prefix is detected, the platform evaluates the subsequent string as JavaScript. While intended as a functional tool for dynamic queries, it effectively turns every unauthenticated query sink into a potential vector for arbitrary script execution.

To mitigate this inherent risk, ServiceNow implemented a "script sandbox" layer. This secondary defense was designed to be restrictive: it blocks eval calls, limits the ability to instantiate sensitive Java classes, prevents the definition of new functions, and restricts unauthenticated users to read-only access for database tables.

Bypassing the Sandbox

The core of the CVE-2026-6875 exploit lies in the researchers’ ability to escape this restrictive sandbox. By manipulating how gs.include() functions interact with the environment, attackers found a way to override global functions. Specifically, by overriding Object.clone with the Function constructor and mapping the AbstractAjaxProcessor.prototype to a malicious payload, an attacker can force the sandbox to execute the Function(payload) command in an unsandboxed context. This grants the attacker the ability to execute arbitrary code with the full privileges of the ServiceNow instance.

Chronology: From Responsible Disclosure to Active Exploitation

The timeline of CVE-2026-6875 underscores the speed at which modern threat actors capitalize on technical disclosures.

• April 1, 2026: Searchlight Cyber officially reports the vulnerability to ServiceNow, adhering to standard responsible disclosure practices.
• April – July 2026: ServiceNow works to develop patches. During this period, the company deploys interim mitigations to cloud-hosted instances, specifically targeting the modification of key JavaScript functions to prevent the identified sandbox escape.
• July 14, 2026: Searchlight Cyber publicly discloses the vulnerability. ServiceNow simultaneously releases patches for all self-hosted instances.
• July 17, 2026: The first reports of in-the-wild exploitation surface. Attackers demonstrate that they are utilizing the /assessment_thanks.do endpoint—the specific sink identified in the PoC—to deliver their payloads.
• July 18, 2026: Threat intelligence firm Defused confirms that while attackers are using the same entry point as the PoC, they have successfully developed alternative "gadget chains" to reach the same code-execution primitive, suggesting that the exploit is being refined and adapted by multiple malicious groups.

Implications: The Scope of the "Blast Radius"

The potential impact of this vulnerability is difficult to overstate. ServiceNow is the central nervous system for many of the world’s largest organizations.

Full System Compromise

Once an attacker gains access via this RCE, the "blast radius" extends far beyond the ServiceNow instance itself. Because the platform is often integrated into the internal corporate network, it acts as a gateway. Attackers can:

1. Exfiltrate Sensitive Data: With unrestricted access to the Rhino engine, malicious actors can query any table within the database, potentially exposing HR records, proprietary project data, or customer PII.
2. Administrative Takeover: Attackers can create new administrator accounts, ensuring persistent access even if the initial vulnerability is patched.
3. Lateral Movement: The report confirms that the exploit allows for shell command execution on proxy servers configured within the internal network. This provides an immediate bridge for attackers to pivot into the internal corporate environment, bypassing perimeter defenses.

Official Responses and Defensive Measures

ServiceNow’s Structural Hardening

In response to the vulnerability, ServiceNow has not merely issued a patch but has fundamentally altered its sandbox architecture. The updated "Guarded Script" mechanism now imposes strict limitations on what can be executed. It prevents variable declarations, control flow statements, function definitions, and complex assignments. By forcing scripts to be limited to a single, simple expression, ServiceNow has significantly raised the bar for future sandbox escape attempts.

Conflicting Advisories

There remains a notable tension between the findings of private security firms and official corporate messaging. While external intelligence firms like Defused have provided concrete evidence of active exploitation, ServiceNow’s official advisory has been slower to update its assessment of the threat landscape, still claiming to be "not currently aware of exploitation" in certain documentation. This discrepancy has led to confusion among IT security teams, many of whom are prioritizing the patch based on the warnings from independent researchers rather than official company communication.

Strategic Recommendations for IT Security Teams

For organizations operating self-hosted ServiceNow instances, the window of opportunity to prevent a breach has effectively closed. If the patches released on July 14 have not been applied, the instance must be considered compromised.

1. Immediate Patching: If a self-hosted instance remains unpatched, it should be taken offline immediately until the security updates are applied.
2. Forensic Investigation: Security teams should conduct a thorough audit of the sysparm_assessable_type parameter in logs. Look for unusual activity around the /assessment_thanks.do endpoint, as this remains the primary vector for attackers.
3. Credential Rotation: If an instance was exposed to the public internet prior to the patch, assume that administrative credentials have been compromised. A full rotation of all service account passwords and administrator credentials is highly recommended.
4. Network Segmentation: Review the role of proxy servers connecting to the ServiceNow instance. Ensure these servers are isolated with strict egress filtering to prevent an attacker from using a compromised ServiceNow instance as a staging ground for lateral movement into the wider internal network.

Conclusion: The Future of Enterprise Security

CVE-2026-6875 serves as a sobering reminder of the inherent risks associated with complex, highly extensible enterprise platforms. The shift from a permissive execution environment to a strictly guarded, expression-based script system is a necessary evolution, but one that highlights the ongoing struggle to balance developer utility with security.

As threat actors continue to evolve their gadget chains and bypass techniques, the security of enterprise software will depend on a shift toward "secure by design" principles. Until then, organizations must rely on rapid patch deployment and the vigilance of independent researchers to identify and close these systemic gaps before they are weaponized against the global digital economy. The rapid exploitation of this flaw demonstrates that for a platform of ServiceNow’s scale, even 72 hours between disclosure and weaponization is a lifetime that many organizations cannot afford.