No longer say AI-assisted cybersecurity—say agentic defense. Google Cloud is taking a proactive stance in its security approach in the era of agentic AI, as evidenced by a series of recent announcements. The acquisition of Wiz also plays a significant role. However, this is a race against time, as malicious actors are launching increasingly sophisticated attacks.
The AI Threat Landscape
As artificial intelligence reshapes usage patterns, it is also profoundly altering the threat landscape. This was the key takeaway from Google Cloud’s Cloud Next ‘26 conference: the AI era demands a complete overhaul of cybersecurity strategies.
Malicious actors are accelerating their operations at an unprecedented pace. According to presented data, cyberattackers now use AI to speed up their operations. The time between an initial intrusion and its takeover by another malicious actor has shrunk from eight hours to just 22 seconds in three years—a shift that requires an equally rapid, machine-scale response.
Beyond this acceleration, a structural transformation is underway. AI introduces new layers to secure—models, agents, data pipelines—while democratizing development: any employee can now create applications or agents. As a result, the attack surface is expanding, rendering traditional methods obsolete.
Cybersecurity Must “Fight AI with AI”
In response to this shift, Google Cloud advocates a clear approach: security must become natively AI-driven. In a world where attackers automate their operations, defense must also be “agentic”—capable of operating continuously at machine speed.
This vision materializes through the launch of new agents in Google Security Operations, designed to detect threats, identify blind spots in detection systems, and enrich analyses with external context. The goal is no longer just to alert but to automate the entire defense cycle.
Google goes further by enabling enterprises to create their own security agents, with remote support for Google Cloud’s Model Context Protocol (MCP) server for Google Security Operations, now available to all.
To streamline the process, enterprises can also access the MCP server client directly from the Google Security Operations chat interface, available in preview. This transforms cybersecurity into a programmable platform tailored to each organization’s specific needs.
Wiz Integration Strengthens the Approach
This automation logic is reinforced by the integration of Wiz, which promotes an end-to-end autonomous security approach. The Israeli-American cybersecurity startup, acquired in March 2025 for $32 billion, was fully integrated this March.
With Wiz’s services, a new model emerges, built on three types of agents:
- Offensive agents simulate real attacks to identify exploitable vulnerabilities.
- Remediation agents automate vulnerability fixes.
- Investigation agents handle incident analysis.
Together, they form a continuous chain capable of identifying, prioritizing, and correcting risks without systematic human intervention. A key advantage lies in their ability to collaborate, solving a long-standing cybersecurity challenge: prioritization. Detected vulnerabilities are no longer theoretical; they are automatically validated, drastically speeding up remediation.
Multi-Cloud and AI: A Systemic Attack Surface
Another major challenge is environment fragmentation. Enterprises now operate in a multi-cloud world, compounded by AI platforms, SaaS tools, and agent studios.
The objective is clear: enable enterprises to apply consistent security policies, regardless of the environment. This is essential as the boundaries between infrastructure, applications, and agents blur.
Context as the Defenders’ Decisive Advantage
In this race for speed, defenders retain a key advantage: context. Unlike attackers, they possess deep knowledge of their systems, critical data, and normal behaviors. Previously difficult to leverage, this context becomes a major strategic asset with AI.
By aggregating this information, defense systems can more accurately detect anomalies, prioritize risks, and anticipate attacks. This capability could become the primary differentiator in environments where attacks are increasingly automated.
Securing the “Agentic Enterprise” and the Future Web
Beyond security operations, Google Cloud is laying the groundwork for comprehensive agent governance with its Gemini Enterprise Agent platform. This introduces several structural components:
- A unique identity for each agent to manage access.
- An Agent Gateway to control interactions between agents and tools.
- Model Armor—now integrated with Agent Gateway, Agent Runtime, Langchain (in preview), and Firebase (available to all)—to defend against new AI-specific threats, such as prompt injections or data leaks.
Similarly, Google is evolving reCAPTCHA into the broader Google Cloud Fraud Defense platform (available to all), capable of distinguishing not only between humans and bots but also between AI agents. This reflects the emergence of an “agentic web,” where automated interactions are becoming the norm.
Security “by Design” from Code to Cloud
Another key focus is securing the AI development lifecycle. The goal is to integrate security from the outset, rather than adding it later.
With Wiz, this includes capabilities such as:
- Analyzing AI-generated code at creation.
- Detecting risky configurations.
- Maintaining a complete inventory of AI tools used within the enterprise to combat shadow AI.
Google is also enhancing its Trusted Cloud with advancements in data protection, network security, and governance. The aim is to provide a secure foundation for increasingly distributed and complex environments.
Toward Cyber Wars at Machine Speed
If there’s one figure to remember, it’s this: “Our triage and investigation agent processed over 5 million alerts in the past year, reducing a typical 30-minute manual analysis to 60 seconds using Gemini,” reports Google. Within 18 to 24 months, the trajectory is clear: cyberattacks and defenses will become largely automated, with “agent vs. agent” confrontations operating continuously.
In this context, the ability to automate all security processes becomes critical. Enterprises must also adopt a proactive stance, using AI to continuously analyze their own vulnerabilities and patch flaws before they are exploited.
Despite escalating threats, there is cautious optimism. The same technologies empowering attackers also provide defenders with unprecedented tools—if deployed rapidly and at scale.
A Key Partnership with Anthropic on Claude Mythos Preview
Speaking of defenders, Google highlighted its participation in Anthropic’s Glasswing project, offering a select group of Google Cloud customers access to Claude Mythos Preview, described by Anthropic as its “most recent and powerful model.”
This partnership aligns with the AI security strategy showcased at Cloud Next ‘26. Mythos is designed as a model evaluation and control layer, identifying risky behaviors—biases, drifts, or exploitable vulnerabilities—before and during deployment.
For Google, the goal is clear: complement runtime protection tools like Model Armor with solutions capable of auditing models in depth. This reinforces a “security by design” vision, where the reliability and robustness of AI systems are verified upfront, not just monitored in production—a critical shift as agents and models become central to critical infrastructures.
A Strategic Battle for Cloud Platforms
Beyond technical announcements, Google is sending a clear strategic signal: cybersecurity is becoming a central pillar of competition in cloud and AI. By combining infrastructure, threat intelligence, autonomous agents, and a multi-cloud ecosystem, the company aims to establish an integrated, open approach that extends far beyond its own environment.
With the strengthened partnership with Wiz and a focus on interoperability, Google positions itself as a cross-platform security layer, suited to a world where enterprises use multiple clouds and platforms simultaneously.
In this new era, cybersecurity is no longer just about protecting systems—it is a key differentiator and potentially a decisive competitive advantage in the race for generative artificial intelligence.
