Google Cloud has announced a wide range of new security features aimed at helping organisations secure their AI projects while also using AI to strengthen cyber defence. The updates, revealed at the Google Cloud Security Summit 2025, include tools to protect AI agents, streamline compliance, and enhance security operations with AI-powered automation.
Securing AI ecosystems
As companies expand their use of AI agents across different environments, protecting those systems has become a key concern. Google Cloud’s Security Command Centre already offers features such as Sensitive Data Protection, which can detect and redact sensitive information to prevent data poisoning, and Model Armor, which helps block threats like prompt injection.
The company is now introducing three new capabilities to improve protection for agentic AI in Google Agentspace and Agent Builder. These include expanded AI agent inventory and risk identification, which will allow automated discovery of AI agents and Model Context Protocol (MCP) servers, helping teams spot vulnerabilities and misconfigurations. Model Armor is also being extended to support real-time safeguards against risks such as jailbreaking and data leakage. In addition, Google Cloud is bringing new threat detection tools that use intelligence from Mandiant to identify suspicious behaviour and external threats in AI systems.
AI-driven security operations
Google Cloud also shared its vision for an “agentic security operations centre” (SOC), where AI agents collaborate to automate tasks such as alert triage, investigation, and response. This system is designed to ease the workload on security teams while improving response times.
One of the first tools, the Alert Investigation Agent, is now available in preview. It can autonomously analyse command-line data, build process trees, and provide verdicts on alerts, along with recommendations for next steps. This approach draws on expertise from Mandiant’s frontline analysts and is intended to reduce manual effort while speeding up incident response.
Alongside this, Mandiant is expanding its consulting services to help organisations adopt AI securely. These services include AI risk governance, pre-deployment security assessments, and threat modelling, which can support businesses in building robust frameworks before deploying generative and agentic AI.
Enhancements to cloud security and compliance
Beyond AI-focused features, Google Cloud announced several updates across its wider security portfolio. In Google Unified Security, new SecOps Labs provide early access to AI-powered experiments in detection and response, while new dashboards give teams better visibility of SOAR data. Chrome Enterprise users also gain enhanced browsing security on iOS and Android, including URL filtering and improved reporting capabilities to reduce risks from unapproved sites and shadow AI.
In its Trusted Cloud portfolio, Google introduced Compliance Manager and Data Security Posture Management, both now in preview. These aim to simplify compliance monitoring and governance for sensitive data, with integrations into BigQuery Security Centre for easier oversight. Risk Reports, powered by virtual red team technology, will also help teams quickly identify and address vulnerabilities.
Updates to Identity and Access Management (IAM) include a new role picker that uses Gemini AI to suggest least-privilege access, and re-authentication requirements for highly sensitive actions. Data security is also being strengthened with expanded monitoring across Vertex AI, BigQuery, and CloudSQL, as well as new support in Cloud Key Management System.
Network security improvements include the preview of Cloud NGFW for high-performance AI workloads, and expanded Cloud Armor Enterprise policies that provide centralised protection for new projects in Google Cloud.
Building a secure future for AI
The updates reflect Google Cloud’s broader strategy of making AI both a tool for defence and an asset to protect. By expanding compliance automation, improving data security, and introducing AI-driven security operations, the company aims to provide organisations with a stronger foundation to adopt AI responsibly and securely.
New customers can try Google Cloud’s security and identity products with US$300 in free credits, while all customers have access to over 20 free services.
