Google Cloud today announced a range of new security capabilities designed to protect organizations’ artificial intelligence capabilities while also using AI to strengthen defenses.

The updates, revealed at the Google Cloud Security Summit 2025, span agent security, operations, consulting and core platform services.

Leading the list is an announcement that Google is extending Security Command Center’s AI Protection tools with three new capabilities, with a focus on securing agentic AI. Google’s Security Command Center is Google Cloud’s centralized security and risk management platform designed to help organizations continuously monitor, detect and remediate threats, misconfigurations and compliance issues.

A newly expanded inventory and risk identification system can now automatically discover AI agents and Model Context Protocol servers, surfacing vulnerabilities and misconfigurations. Model Armor, Google Cloud’s in-line protection system that defends AI agents against threats such as prompt injection, data leakage and tool poisoning, is being enhanced with in-line protection for Google Agentspace prompts and responses with the provision of real-time safeguards against prompt injection, jailbreaking and sensitive data leakage.

The third new release for Security Command Center sees new threat detections for AI agents being added, leveraging Mandiant and Google intelligence to spot anomalous behavior and external threats.

Complementing the updates for Security Command Center, Google also outlined its vision for an “agentic SOC.” An Alert Investigation agent in Google Security Operations, now in preview, can autonomously enrich events, analyze command-line activity and build process trees modeled on Mandiant analyst practices. The agent automates alert triage and investigation to reduce manual workloads and speeds response times.

The consulting arm of Google-owned Mandiant is also expanding its role in AI security with new services for governance frameworks, AI environment hardening and threat modeling. The expansion is aimed at helping organizations assess risk through adversarial testing and implement comprehensive defense mechanisms ahead of deploying generative or agentic AI.

Google’s Unified Security Platform is also receiving updates that combine capabilities across its ecosystem, infused with Gemini AI. Within Google Security Operations, a new SecOps Labs provides early access to experimental AI-driven features for parsing, detection and response, while new dashboards, now generally available, integrate security orchestration, automation and response data to improve visibility. Chrome Enterprise is extending protections to iOS and Android, with URL filtering and enhanced reporting aimed at reducing shadow AI risks.

On the cloud platform side, Security Command Center is adding several previews: Compliance Manager to unify compliance workflows; Data Security Posture Management integrated with BigQuery Security Center; and new risk reports powered by virtual red team technology.

Identity and Access Management will soon support “Agentic IAM” for provisioning agent identities, a Gemini-powered role picker to recommend least-privilege roles and re-authentication for sensitive actions, enabled by default but with the option to opt out.

Data security enhancements include expanded coverage in Sensitive Data Protection and new capabilities in Cloud Key Management System Autokey, now generally available and network security updates include org-scope tagging in Cloud NGFW, support for RDMA networks in preview and expanded Cloud Armor Enterprise features such as hierarchical policies and ASN-based threat intelligence.

Image: SiliconANGLE/Reve