Google has unveiled new AI security capabilities that include expanded inventory and risk detection for its Agentspace and Agent Builder platforms, featuring layered protections such as detective policies, preventative controls, and tools like Model Armor accessible via API for tailored security.

These updates also introduce automated discovery mechanisms to monitor and prevent unauthorized AI access or manipulation, addressing rising threats like poisoning and prompt injections, with support for the Model Context Protocol (MCP) in Security Command Center to enhance vulnerability detection.

The new security features aim to bolster defenses against AI-specific risks, including tool poisoning and prompt injection, while integrating with Security Command Center to improve misconfiguration detection and vulnerability management.

These advancements support the development of an agentic security operations center (Agentic SOC), where AI agents coordinate actions to improve security team efficiency and automate threat response.

Google has enhanced its Chronicle Security Operations platform with AI-powered autonomous 'agentic SOCs' capable of investigating alerts, analyzing data, and recommending responses, significantly reducing manual workload.

These AI-driven security tools are designed to accelerate threat response times from hours to minutes, which is crucial for preventing costly breaches and regulatory penalties.

Identity and Access Management (IAM) has been upgraded with features like Agentic IAM for auto-provisioning, a role picker with Gemini chatbot for least-privilege access, and re-authentication for sensitive actions, alongside expanded monitoring of AI assets.

Security Command Center now includes features such as Compliance Manager, Data Security Posture Management, and Risk Reports, all in preview, to enhance policy compliance, sensitive data governance, and security issue summaries.

Chrome Enterprise has extended security protections to Android and iOS devices, including URL filtering and improved reporting, to reduce shadow AI risks and monitor security events across mobile platforms.

Google Security Operations has introduced new AI-powered dashboards and SecOps Labs for better threat detection, parsing, and response, integrating security data for operational efficiency.

Support for the Model Context Protocol (MCP) within Vertex AI enhances secure communication and mitigates risks like prompt injection, with applications in sectors such as financial trading to safeguard algorithms.

Google emphasizes securing the entire AI development pipeline to protect intellectual property, ensure compliance, and foster trust, highlighting that security should facilitate innovation.

Model Armor technology now extends real-time protection to Agentspace, defending against threats like prompt injection, jailbreaking, and data leaks during agent interactions.