Google is introducing a range of new security capabilities, with AI agents at the center. At the core is the integration of the acquired Wiz.
Following the acquisition, Wiz will remain available for all major cloud platforms, including AWS, Azure, and Oracle Cloud. With the integration of Wiz, Google aims to build a platform that simplifies multi-cloud security. To that end, Wiz is introducing an AI Application Protection Platform (AI-APP) and security agents for red-team, blue-team, and green-team tasks. Allowing Wiz to operate independently across cloud boundaries was an essential condition during negotiations—in 2024, Wiz rejected an initial offer of $23 billion.
M-Trends 2026: 22 seconds to respond
Google subsidiary Mandiant is publishing M-Trends 2026 today, based on more than 500,000 hours of incident investigations. It shows that cybercriminals are increasingly operating like organized enterprises. They have reduced the window for defenders to intervene from several hours to 22 seconds. Attackers are not only targeting data theft but are also attempting to completely dismantle organizations’ resilience to maximize extortion pressure.
In addition, Mandiant is publishing a separate report on AI risks based on 2025 data. The research shows that over the past year, attackers have shifted from experimenting with AI to deploying autonomous tools that rewrite their own code in real time. Mandiant advises organizations to replace passive governance with continuous red teaming of AI models and agents.
Agentic SOC and dark web intelligence
Google Security Operations is introducing agentic automation, now available as a preview. Security teams can embed a Triage and Investigation agent directly into their workflows. This agent independently investigates alerts, gathers evidence, and delivers conclusions with accompanying explanations.
Also new is dark web intelligence in Google Threat Intelligence, powered by the latest Gemini models. The feature automatically builds an organization profile and analyzes millions of external signals daily with 98 percent accuracy.
For AI security, Security Command Center integrates with Vertex AI Agent Engine, while Model Armor now also supports Google MCP servers to mitigate prompt injection and data breaches. Remote MCP server support in Google Security Operations will be generally available in early April.
Tip: What are Google’s plans for cloud security provider Wiz?