AI Security AI安全 14h ago Updated 2h ago 更新于 2小时前 49

Critical ServiceNow AI Platform Flaw Exploited for Unauthenticated Code Execution ServiceNow AI平台关键漏洞被利用实现未授权代码执行

A critical sandbox escape vulnerability (CVE-2026-6875, CVSS 9.5) in the ServiceNow AI Platform is being actively exploited in the wild to achieve unauthenticated code execution. The flaw allows attackers to completely compromise ServiceNow instances and connected proxy servers via the "/assessment_thanks.do" endpoint using HTTP POST requests. ServiceNow released patches across multiple versions (Brazil, Australia, Zurich, Yokohama) in June and is implementing stricter restrictions on code execu 威胁情报机构Defused Cyber报告正在野外利用ServiceNow AI平台的关键漏洞CVE-2026-6875进行未授权代码执行攻击。 该漏洞CVSS评分高达9.5,属于沙箱逃逸漏洞,允许未经身份验证的用户在实例及关联代理服务器上运行任意代码。 ServiceNow已在多个版本(如Brazil, Australia, Zurich, Yokohama)中发布补丁,并计划通过限制沙箱上下文中的代码类型来增强安全性。 尽管第三方声称观察到利用活动,ServiceNow官方回应称在其托管实例中尚未发现相关证据,但仍强烈建议所有客户应用补丁。

75
Hot 热度
65
Quality 质量
70
Impact 影响力

Analysis 深度分析

TL;DR

  • A critical sandbox escape vulnerability (CVE-2026-6875, CVSS 9.5) in the ServiceNow AI Platform is being actively exploited in the wild to achieve unauthenticated code execution.
  • The flaw allows attackers to completely compromise ServiceNow instances and connected proxy servers via the "/assessment_thanks.do" endpoint using HTTP POST requests.
  • ServiceNow released patches across multiple versions (Brazil, Australia, Zurich, Yokohama) in June and is implementing stricter restrictions on code execution within sandbox contexts.
  • While threat intelligence firms report active exploitation, ServiceNow officially stated they have not observed evidence of attacks targeting their hosted instances, urging all customers to apply available fixes immediately.

Why It Matters

This incident highlights the severe risks associated with enterprise AI platforms that integrate complex scripting and sandboxing mechanisms, where a single misconfiguration can lead to full system compromise. For security practitioners and IT administrators, it underscores the urgency of patching critical vulnerabilities in widely used workflow automation tools, as unauthenticated access can bypass traditional perimeter defenses. The discrepancy between independent threat intelligence reports and vendor statements also emphasizes the need for proactive verification and rapid response in enterprise security operations.

Technical Details

  • Vulnerability Type: Sandbox escape leading to arbitrary code execution, rated Critical with a CVSS score of 9.5.
  • Attack Vector: Exploitation targets the pre-authentication endpoint "/assessment_thanks.do" via HTTP POST requests, allowing unauthenticated users to execute malicious payloads.
  • Impact: Successful exploitation results in a complete compromise of the ServiceNow instance and all connected proxy servers, granting attackers full control over the environment.
  • Remediation: Patches were distributed in June for various platform versions including Brazil EA/GA, Australia Patch 2, Zurich Patch 7b/9, and Yokohama Patch 12/13. ServiceNow is also enhancing security by restricting permissible code types in sandbox contexts.
  • Disclosure Timeline: Reported by Searchlight Cyber on April 1, 2026, with in-the-wild exploitation confirmed by Defused Cyber shortly thereafter.

Industry Insight

Enterprise organizations must prioritize immediate patch management for AI-enabled workflow platforms, treating sandbox vulnerabilities with the same severity as direct server breaches. The presence of active exploitation campaigns suggests that threat actors are specifically targeting AI integration points, necessitating enhanced monitoring of API endpoints and sandbox execution logs. Furthermore, vendors should maintain transparent communication regarding threat levels, as conflicting information between independent researchers and official statements can hinder effective incident response planning.

TL;DR

  • 威胁情报机构Defused Cyber报告正在野外利用ServiceNow AI平台的关键漏洞CVE-2026-6875进行未授权代码执行攻击。
  • 该漏洞CVSS评分高达9.5,属于沙箱逃逸漏洞,允许未经身份验证的用户在实例及关联代理服务器上运行任意代码。
  • ServiceNow已在多个版本(如Brazil, Australia, Zurich, Yokohama)中发布补丁,并计划通过限制沙箱上下文中的代码类型来增强安全性。
  • 尽管第三方声称观察到利用活动,ServiceNow官方回应称在其托管实例中尚未发现相关证据,但仍强烈建议所有客户应用补丁。

为什么值得看

对于依赖ServiceNow进行企业流程自动化和AI集成的组织而言,此事件揭示了关键基础设施中未授权访问的严重风险,强调了及时修补高危漏洞的紧迫性。同时,厂商与独立安全研究者在威胁观察上的差异也提醒从业者需保持多方情报验证的习惯,以准确评估自身环境的安全态势。

技术解析

  • 漏洞详情:CVE-2026-6875是一个严重的沙箱逃逸漏洞,允许攻击者绕过认证机制执行任意代码,导致整个ServiceNow实例及连接的代理服务器完全被控制。
  • 受影响版本:补丁覆盖范围包括Brazil EA/GA、Australia Patch 2、Zurich Patch 7b/9以及Yokohama Patch 12 Hot Fix 1b/13。
  • 攻击向量:初始报告指出攻击针对预认证端点/assessment_thanks.do使用HTTP POST请求,后续更正指出实际载荷匹配Searchlight Cyber发布的PoC,通过不同的沙箱逃逸 gadget 路径实现相同的代码执行原语。
  • 缓解措施:ServiceNow不仅发布了修复补丁,还宣布将严格限制可在沙箱上下文中运行的代码类型,从架构层面降低类似逃逸攻击的成功率。

行业启示

  • 零信任与补丁管理:即使是大厂级的企业服务平台,其AI集成模块也可能成为攻击入口,企业必须建立快速响应机制,确保高危漏洞(CVSS > 9.0)得到即时修补。
  • 情报交叉验证:第三方安全公司、漏洞披露者和厂商官方之间关于“是否正在被利用”的信息存在不一致,安全团队应结合多种来源的情报,而非单一依赖某一方声明,以制定更稳健的防御策略。
  • 云与本地部署的风险差异:虽然ServiceNow声称其托管实例未受影响,但自托管版本用户面临直接威胁,这凸显了不同部署模式下安全责任边界和风险暴露面的差异,需针对性加强监控。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Security 安全