Google Cloud Fortifies Network Security for the Agentic AI Era

Google Cloud has unveiled a comprehensive set of new security capabilities designed to bolster the protection of its platform and the artificial intelligence (AI) technologies it hosts. These advancements, announced at the Google Cloud Security Summit 2025, signal a strategic focus on securing AI innovation while simultaneously harnessing AI to enhance overall enterprise security postures. The updates span critical areas including AI agent security, security operations, expert consulting services, and core platform enhancements, reflecting a proactive approach to the evolving threat landscape driven by AI.

Securing the AI Ecosystem

A primary focus of the new announcements is the enhanced security for AI agents. Google Cloud is extending its Security Command Center's (SCC) AI Protection tools with three key capabilities. These include an expanded AI agent inventory and risk identification system, which can automatically discover AI agents and Model Context Protocol (MCP) servers, thereby surfacing potential vulnerabilities, misconfigurations, and high-risk interactions. This provides security teams with greater visibility and control over their AI agent ecosystem. Additionally, Google Cloud is enhancing Model Armor, its in-line protection system designed to defend AI agents against threats like prompt injection, data leakage, and tool poisoning. Model Armor will now offer in-line protection for prompts and responses within Google Agentspace, providing real-time safeguards against runtime threats such as prompt injection, jailbreaking, and sensitive data leakage. Complementing these measures, new threat detections for AI agents are being integrated into SCC, leveraging intelligence from Mandiant and Google to identify anomalous behavior and external threats, thereby improving incident response capabilities for risky behaviors on agent assets.

The Agentic Security Operations Center (SOC) Vision

Google Cloud is also advancing its vision for an "agentic SOC," a paradigm where AI agents coordinate their actions to achieve shared security objectives. This involves integrating AI agents into security operations to automate critical tasks such as alert triage, investigation, and response. A key component of this vision is the Alert Investigation agent, currently in preview within Google Security Operations. This agent is designed to autonomously enrich events, analyze command-line activity, and build process trees based on the best practices of Mandiant's frontline analysts. By automating these tasks, the agent aims to significantly reduce manual workloads for security teams and accelerate response times. The agent not only assesses alerts but also provides recommendations for human intervention, allowing analysts to focus on more complex investigations and strategic threat hunting.

Mandiant Consulting Enhances AI Security Expertise

In response to the growing demand for guidance on securing AI applications, Mandiant Consulting, Google Cloud's cybersecurity consulting arm, is expanding its AI security services. These new offerings include risk-based AI governance, pre-deployment guidance for AI environment hardening, and AI threat modeling. By partnering with Mandiant, organizations can gain expert assistance in assessing risks through adversarial testing and implementing comprehensive defense mechanisms before deploying generative or agentic AI. This expansion reflects Mandiant's commitment to helping organizations navigate the complexities of AI security and mitigate associated risks.

Google Unified Security Platform Updates

The Google Unified Security (GUS) platform is also receiving significant updates, integrating capabilities across Google Cloud's security ecosystem and infused with Gemini AI. Within Google Security Operations, the new SecOps Labs provides early access to experimental AI-driven features for parsing, detection, and response, allowing security teams to explore and leverage cutting-edge AI tools. Additionally, new dashboards in Google Security Operations, now generally available, offer a native integration of Security Orchestration, Automation, and Response (SOAR) data, enhancing visibility and enabling users to visualize, analyze, and act on their security data more effectively. Chrome Enterprise is extending its security protections to iOS and Android, including URL filtering and enhanced reporting capabilities, aimed at reducing shadow AI risks and strengthening Google's position in the enterprise browser market.

Expanded Cloud Security Tools and Identity Management

Beyond AI-specific enhancements, Google Cloud is introducing several updates to its broader cloud security portfolio. Security Command Center is being updated with Compliance Manager, which unifies policy definition, control configuration, enforcement, monitoring, and evidence generation for auditing, streamlining security, privacy, and compliance workflows. New recommended AI controls within Compliance Manager automate AI workload compliance through built-in baselines and continuous monitoring. Data Security Posture Management (DSPM), also in preview, provides governance for the security and compliance of sensitive data, with a new native integration with BigQuery Security Center allowing data professionals to monitor their data security posture directly within the BigQuery console. Risk Reports, powered by SCC's virtual red team technology, highlight unique cloud security issues that could put an organization at risk of attack, enabling quick identification and remediation of defense gaps. In the realm of Identity and Access Management (IAM), the forthcoming Agentic IAM service will enable the provisioning of agent identities across cloud environments, with Google Cloud auto-provisioning these identities and supporting various credential types and authorization policies. The preview of the IAM role picker, powered by Gemini, allows for the granting of least-privilege access by describing desired actions, with Gemini recommending the most appropriate roles. Re-authentication for sensitive actions, also in preview, will trigger re-authentication for critical actions like updating billing accounts, enhancing protection against unauthorized access.

Data Security and Network Security Enhancements

Data security measures are also being strengthened. Sensitive Data Protection is being expanded to monitor Vertex AI, AI-related assets in BigQuery and CloudSQL, inspect images for sensitive data such as barcodes and license plate numbers, and detect AI and machine learning context models for various data types including medical records and financial invoices. Cloud Key Management System Autokey in Cloud Setup is now generally available, assisting customers with customer-owned encryption keys to onboard more quickly with recommended key management practices. Network security is being bolstered with expanded org-scope tags in Cloud NGFW, providing hierarchical support, and preview support for Zero Trust networking in high-performance computing VPCs, including AI workloads. Cloud Armor Enterprise now offers generally available support for hierarchical security policies and organization-scoped address groups, facilitating centralized control and automatic protection of new projects. Updates to Cloud Armor also include enhanced inspection limits for WAF rules and added support for rate limiting by JA4 fingerprints and ASN-based threat intelligence for Media CDN.

Empowering Defenders and Responsible AI Adoption

These innovations collectively demonstrate Google Cloud's commitment to making security an enabler of business ambitions. By automating compliance, simplifying access management, and expanding data protection for AI workloads, organizations can enhance their security posture with greater speed and ease. The company emphasizes that by using AI to empower defenders and meticulously securing AI projects from inception to deployment, Google Cloud provides a comprehensive foundation for organizations to thrive in the AI era. The updates aim to provide a layered set of built-in protections, making the secure path the easy path without sacrificing developer flexibility, offering both detective policies for visibility and preventative controls to block high-risk actions by default.