Las Vegas ― At Google Cloud Next 25, the tech giant unveiled a sweeping vision for the future of cybersecurity, tightly woven with the fabric of artificial intelligence. At the forefront of this transformation are two major innovations: Security Agents and Google Unified Security (GUS). These technologies exemplify Google’s mission to reimagine the role of AI in safeguarding data, infrastructure, and digital experiences at scale.
Security Agents represent a breakthrough in incident response and threat management. These AI-powered agents are designed to analyze malware, triage alerts, and accelerate the pace of investigations. Operating with proactive intelligence, they reduce the burden on security teams by automating complex detection workflows and correlating high-risk behaviors across sprawling IT environments.
Already in deployment across major organizations, Security Agents are not just reactive tools, but tireless defenders capable of learning from every encounter to become more efficient with each new threat. Their integration into platforms like Google SecOps and Google’s Security Command Center is already yielding results—helping clients like Charles Schwab and Dun & Bradstreet cut resolution times and boost visibility across systems.
Taking this integration further is Google Unified Security (GUS), a comprehensive and converged platform uniting detection, investigation, and automated response under one AI-powered roof. GUS offers unmatched visibility across cloud environments, endpoints, and networks—whether in development stages or full deployment.
Mandiant’s expertise woven into Google Unified Security
What sets GUS apart is its ability to apply continuous virtual red teaming, rapidly identify data leaks, and enact instant remediation via AI agents, all while operating at a planetary data scale. Through real-time dashboards and Gemini-enhanced intelligence, GUS not only identifies threats but understands intent, assesses exposure, and initiates automated countermeasures within seconds. Developers and security analysts alike benefit from this ecosystem where previously siloed tools now act in unison, increasing both coverage and agility.
One illustrative scenario at Next ’25 involved a developer unknowingly installing a risky Chrome extension. GUS flagged the data leak, isolated the threat, and recommended strengthening the AI model using Google’s Model Armor—all within minutes and with minimal human input. Such capabilities demonstrate not just the promise of AI in cybersecurity, but its active readiness today.
A key element reinforcing the strength of Google’s security ecosystem is the integration of Mandiant’s frontline expertise into its AI-driven solutions. Acquired by Google Cloud in 2022, Mandiant brings decades of real-world threat intelligence, incident response experience, and adversary behavior profiling. This deep knowledge has now been woven directly into GUS, elevating its capacity to respond not just automatically—but insightfully.
Mandiant’s influence is evident in features like continuous virtual red teaming and high-fidelity threat detection, which are designed to emulate real-world attacker behavior and uncover vulnerabilities traditional tools may overlook. Additionally, through Mandiant Cybersecurity Consulting, enterprises gain access to tailored security guidance from seasoned experts who have managed high-stakes incidents across global sectors. These consultants work alongside customers to assess risk, build resilient security architectures, and ensure AI-driven tools like Google SecOps and GUS are effectively aligned with operational needs.
By fusing Mandiant’s battle-tested intelligence and hands-on consulting with Google’s AI engines and infrastructure, GUS is positioned not just as a reactive tool, but as a predictive and strategic guardian of enterprise systems.
Reimagine through AI: Collaboration between Sphere and Google Cloud
Yet Google’s ambitions extend far beyond the perimeter of traditional IT security. The company also used its keynote stage to spotlight how AI is reshaping industries from entertainment to infrastructure. One of the most talked-about announcements was Google’s collaboration with Sphere in Las Vegas, where a technologically reimagined production of The Wizard of Oz is set to debut on August 28, 2025. This immersive theatrical experience is being brought to life through a fusion of Google Cloud’s scalable AI infrastructure and cutting-edge creative tools from Google DeepMind.
The project is both a technical marvel and a nostalgic homage. Through the use of fine-tuned Gemini models, Veo 2, and Imagen 3, the production team is not only enhancing the resolution and extending the backgrounds of the film, but also digitally recreating characters that could not otherwise appear on the same screen—an ambitious feat reminiscent of the original film’s pioneering use of Technicolor nearly a century ago. To support this, Google Cloud has handled over 1.2 petabytes of data throughout development, showcasing the computational depth behind what appears on screen for just a few minutes.
This AI-powered storytelling doesn’t only represent a leap in entertainment but symbolizes a broader shift toward immersive, intelligent digital experiences. It highlights how AI, when matched with creative intent and massive-scale infrastructure, can reimagine even the most iconic pieces of culture.
Enhancements to chip and AI model: Ironwood, Gemini 2.5
Building on this spirit of innovation, Google CEO Sundar Pichai took the stage to introduce some of the most ambitious technological leaps yet. Chief among them was Ironwood, Google’s seventh-generation Tensor Processing Unit (TPU), representing a staggering 3,600-fold increase in performance compared to its first publicly available TPU. Designed to meet the surging computational demands of large-scale AI workloads, Ironwood is not just faster but dramatically more energy-efficient—29 times more than earlier generations—making it a foundational piece for the next era of AI infrastructure. It’s this kind of raw processing power that enables breakthroughs in fields ranging from quantum computing to global weather forecasting.
But it was the unveiling of Gemini 2.5, Google’s most advanced AI model to date, that truly captured the imagination. Unlike previous iterations, Gemini 2.5 is a “thinking model”—capable of reasoning through complex problems before generating a response. This deeper cognitive capacity allows it to excel in highly abstract tasks, including solving dynamic simulations like a Rubik’s Cube or rendering real-time physics models of the Earth’s magnetic field and general relativity. According to industry benchmarks such as Chatbot Arena, Gemini 2.5 currently ranks as the most intelligent model in the world, even earning the highest score ever recorded on “HumanEval,” a rigorous test meant to reflect the boundaries of human reasoning.
In addition, Google announced Gemini 2.5 Flash, a cost-efficient, low-latency variant that allows users to fine-tune the model’s reasoning depth to balance performance with budget constraints. This flexibility ensures that businesses of all sizes can benefit from cutting-edge AI, whether they’re running lightweight applications or complex enterprise workflows.
With this full-stack investment—from custom silicon and quantum chips to frontier models and globally distributed infrastructure—Google is signaling not only its leadership in the AI race but also its commitment to democratizing access to these transformative tools. As Pichai noted, the goal is to “organize the world’s information and make it universally accessible and useful”—a mission that AI, in its most powerful and responsible form, is now helping to fulfill at unprecedented scale.
