Anand Oswal, EVP – Products, Palo Alto Networks.
If the story of enterprise technology were to be told, the tale would begin placidly with mainframes and simple packet-switch networks. Minor ripples such as the shift from on-prem to cloud and the evolution to SD-WAN networks will be pivotal points of interest in the story.
However, the AI transition we’re in the midst of would be the technology equivalent of the Titanomachy in Greek mythology—a complete shift in the world order. AI is now becoming the digital surface itself. As organizations move beyond simple chatbots toward autonomous “compound systems” of agents (and with the evolution of frontier models such as Mythos and GPT-5), the traditional tech landscape has shifted.
The Lid Is Off
To continue with the Greek myth reference, there’s an often-told story of Pandora opening a storage jar. Once the lid was lifted, the challenges became part of the atmosphere itself.
Today’s enterprise faces a similar “Pandora moment.” As we integrate AI models into our core architecture, build agents and use agentic browsers, we’ve lifted the lid on autonomous agency. We can’t put the agents back in the jar; our task now is to govern the world they’ve already begun to reshape.
AI is now embedded in four primary surfaces: applications, autonomous agents, IT-managed employee laptops and the browser. To navigate this shift, leaders must recognize that as AI’s capabilities expand, the attack surface mutates.
The Flip Side Of The AI Evolution
For decades, application architecture was predictable. Today, the “AI stack” has introduced a complex web of models, plug-ins, vector databases and real-time internet search integrations. These agents create an exponential increase in supply chain vulnerabilities.
We’re moving rapidly toward an era of “mass agency.” An October 2025 Gartner, Inc. report found that 81% of surveyed enterprises were piloting the use of AI agents or had fully implemented AI agent solutions. These agents are taking actions—filing tickets, orchestrating workflows and accessing sensitive data.
Much like the spirits released by Pandora, these risks such as identity compromise, sensitive data leak, tool misuse, memory poisoning and indirect prompt injection move with a speed and autonomy we’ve never seen. When agents begin to communicate via protocols such as Model Context Protocol (MCP), new attack vectors emerge that don’t exist at the AI application level.
The Frontiers Of ‘Vibe Coding’ And Agentic Browsers
The rise of “vibe coding,” where developers use natural language to generate entire codebases, has unlocked massive productivity. However, for that agent to “vibe,” it needs local system access. It becomes a potential Trojan horse; one prompt injection from a public repository could exfiltrate proprietary source code before a human even reviews the logs.
Simultaneously, the browser has become the de facto workspace. A 2025 Palo Alto Networks report noted that 85% of daily work was happening inside a tab. As we move toward agentic browsers, every tab becomes a 24/7 assistant with access to cookies and credentials. Without a unified security strategy, the browser becomes the ultimate blind spot for governance.
Beyond Hope: Fighting AI With AI
When Pandora closed the lid on the jar, only “Hope” remained at the bottom. In the modern AI enterprise, we know hope isn’t a strategy. We can’t hope that legacy firewalls will catch a prompt injection or hope that developers aren’t bypassing identity controls.
To stay ahead of AI-powered attacks, we must transform our defenses to fight AI with AI. The reality is that the same frontier models that find and exploit vulnerabilities can also be part of the defense, but only if they’re quickly integrated into defensive solutions.
Real-time, AI-driven, unified platforms can make this speed possible. Because AI feeds on data, it’s critical to scaffold together data and context. This doesn’t mean you must have a single platform, but context can’t be spread across hundreds of different products and vendors if you need to move at machine speed.
The Golden Age Of Modern Security Architecture
AI and autonomous agents represent the greatest leap forward in enterprise and personal productivity and capability that we’ve ever seen. To safely unlock this potential, we must secure every AI interaction. We need security to be built into your product’s architecture by design.
The goal is no longer to put the evils back in the box; it’s to out-compute them. By moving from passive hope to an active, open and AI-driven defense, organizations can finally harness the power of the agentic revolution with confidence.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?