Google Cloud Next '26: The $750M Bet on Agentic AI Sparks a New Security Arms Race

At Google Cloud Next '26, the company made a bold declaration: it intends to become the operating system for enterprise AI. With a $750 million fund dedicated to agentic AI, new custom TPU chips designed for faster and more efficient AI workloads, and the launch of the Gemini Enterprise Agent Platform, Google is betting big on a future where autonomous agents handle complex business tasks. However, this rapid acceleration into agentic AI creates a new security frontier that demands immediate attention from the cybersecurity community.

The announcements were staggering in scope. The $750 million fund is designed to support partners and enterprises building AI agents, providing access to Gemini models and infrastructure credits. The new TPU chips, Google's latest generation, promise to make AI faster and more efficient, enabling real-time agent interactions. The Gemini Enterprise Agent Platform allows companies to deploy, manage, and scale AI agents across their operations, from customer service to supply chain management.

But with great power comes great risk. AI agents, by their nature, are autonomous. They make decisions, execute actions, and interact with other systems without direct human oversight. This autonomy creates a fundamentally new attack surface. Traditional security tools are ill-equipped to monitor and protect these agents. Recognizing this, Google's ecosystem partners have stepped up with specialized solutions.

Check Point announced the integration of its AI Defense Plane with Google Cloud. This solution provides integrated discovery, governance, and runtime protection for AI agents. It can identify all agents operating within an environment, enforce policies on what data they can access, and monitor their behavior in real-time to detect anomalies. This is crucial because an agent compromised by a prompt injection attack could exfiltrate sensitive data or execute unauthorized transactions.

Exabeam extended its Agent Behaviour Analytics to the Google Cloud Agent Ecosystem. By applying its expertise in user and entity behavior analytics (UEBA) to AI agents, Exabeam can establish baselines for normal agent behavior and flag deviations. This is particularly important for detecting subtle attacks, such as an agent slowly escalating its privileges or communicating with a command-and-control server.

Rubrik took a different approach, focusing on securing and accelerating AI agents through data protection. Rubrik's platform ensures that the data used and generated by AI agents is backed up, recoverable, and immutable. This is vital because if an agent is compromised or makes a catastrophic error, the organization needs to be able to roll back to a known good state. Rubrik's integration with Google Cloud provides a safety net for agent operations.

For cybersecurity professionals, the message from Google Cloud Next '26 is clear: agentic AI is not coming; it is here. The infrastructure investments and new tools signal a paradigm shift. The security industry must move beyond traditional perimeter defenses and adopt a zero-trust model for AI agents. This means verifying every action an agent takes, continuously monitoring its behavior, and having robust incident response plans for agent-related incidents.

The $750 million fund will likely accelerate the development of both legitimate agents and, inevitably, malicious ones. Threat actors will also have access to these tools and models. The security community must be proactive, not reactive. The partnerships announced at Next '26 provide a starting point, but enterprises need to build internal capabilities to manage this new risk.

In conclusion, Google Cloud Next '26 marked the beginning of a new era. The race to deploy agentic AI is on, and so is the arms race to secure it. For those in cybersecurity, this is both a challenge and an opportunity. The tools and strategies we develop today will define the security posture of the AI-powered enterprise of tomorrow.