Partner POV | The 2026 AI Frontier: Securing the New Enterprise Architecture - A Collaborative Perspective from WWT & Cloudflare

This article was written and contributed by, Cloudflare.

Summary

Cloudflare, in partnership with WWT, offers a unified platform for securing workforce AI tools and public-facing applications. By discovering shadow AI, protecting models from abuse, securing agent access, and preventing data exposure in prompts, organizations can innovate safely and efficiently with easier visibility and stronger control.

The 2026 AI Frontier

In 2026 and beyond, the conversation around artificial intelligence has shifted from experimental pilots to foundational enterprise architecture. According to WWT Research, AI is no longer just an "add-on" but a force that will "change everything" about how businesses operate. However, this shift introduces a paradox: the faster an organization innovates, the larger its "shadow AI" footprint and security attack surface become.

To navigate the paradox of faster AI adoption and safer AI operations, business leaders must rethink their approaches to both infrastructure architectures and security best practices.  Traditional approaches to architecture and security when it comes to AI may result in failures on both fronts.  Luckily, WWT and Cloudflare can help organizations adopt new AI architectures and AI governance strategies to accelerate their AI safety and success.

Based on WWT extensive research into 2026 priorities around AI, here are four important areas IT leaders should focus on when it comes to making AI safer and more successful in their organizations, and how Cloudflare can help organizations achieve them.

1. The architectural shift: Moving from AI silos to fabric

WWT's research on Rethinking Enterprise Architecture emphasizes that AI must be treated as a holistic data-and-compute fabric rather than a series of disconnected tools. By the end of 2026, the goal for organizations is successful "agentic AI"—systems that don't just answer questions but execute multi-step workflows. As AI's capabilities will expand over time, expect to see easier and more robust levels of autonomy, agency, and inter-agent collaboration which can lead to increased risks. 

WWT's view: Successful AI integration, especially agentic AI, requires a specialized high-performance architecture that prioritizes low latency and high data mobility. WWT notes that deploying AI agents may also expose hidden data problems, creating compliance risks if AI data governance is lacking.

How Cloudflare can help: Cloudflare's AI Gateway serves as the unified control plane for this new, high-performance fabric. As WWT's architecture moves data across hybrid clouds, Cloudflare ensures that every request to an LLM (whether OpenAI, Anthropic, or internal models) is logged, cached for performance, and rate-limited to control costs. Cloudflare AI Gateway also enforces content safety guidelines by identifying and blocking or redacting harmful content and personally identifiable information from prompts and responses for a better compliance and governance posture.

2. Combating "shadow AI" and the 2026 threat landscape

A recurring theme in WWT's AI and Data Priorities for 2026  and Security Priorities for 2026 is the loss of visibility and the increased risk from AI. The truth for most organizations is that employees are adopting GenAI tools faster than IT can vet them—a phenomenon Cloudflare refers to as shadow AI.

• The problem: WWT notes that sensitive corporate IP is increasingly "leaking" into public models through well-intentioned employees using shadow AI, exposing organizations to increased risks.
• The defense: Cloudflare's AI Security Suite addresses this via Cloudflare SASE (Secure Access Service Edge). It allows organizations to:
  • Discover: Automatically identify which AI apps are being used across the network.
  • Control: Implement zero trust policies to block unapproved tools while allowing sanctioned ones like ChatGPT Enterprise.
  • Protect: Use AI-powered DLP (Data Loss Prevention) to redact PII (Personally Identifiable Information) or source code from prompts in real-time before they reach the provider.

3. Hardening the "front door": Firewall for AI

While much of the security focus around AI is on internal use and misuse, both WWT and Cloudflare believe organizations need to address the risks of public-facing AI applications. If organizations build an AI chatbot or deploy an agent for customers, that chatbot or agent is a new vector for attack that will be exploited if not properly secured.

WWT's research highlights that operationalizing AI-specific threat management and incident response should be a top priority in 2026. Attackers are moving beyond simple hacking to more sophisticated methods specifically targeting AI such "prompt injection" and "model poisoning."

Cloudflare's Firewall for AI provides a model-agnostic layer of defense that sits in front of these AI applications. It identifies:

• Malicious prompts by detecting "jailbreak" attempts designed to make the AI ignore its safety guidelines.
• Toxicity & compliance by using integrated models like Llama Guard to filter out unsafe or non-compliant responses before the customer sees them.

4. Strategic priority: Cost-effective connectivity

One of the most significant insights from WWT's research is the need to optimize architectures and data to fit an organization's AI initiatives—including AI workload and data placement.  This is especially true when organizations must move massive datasets between clouds for training and inference purposes—where data egress fees may put financial hurdles in front of AI innovation. 

Cloudflare's R2 Storage (part of Cloudflare connectivity cloud) enables WWT's vision of a "fluid" enterprise architecture by eliminating egress fees. This allows organizations to store large amounts of training data in one place and run inference wherever the compute is most efficient, ensuring that security and cost-efficiency scale alongside the AI's intelligence.

A checklist for 2026 AI readiness

While AI adoption will cause IT and business leaders to re-evaluate their infrastructure, security, and governance frameworks, it is often difficult to know what to immediately focus on.  Based on the combined insights of WWT and Cloudflare, enterprise leaders seeking to capitalize and accelerate their AI strategies for tangible results in 2026 should prioritize:

1. Visibility first: Use Cloudflare's AI Gateway to see what models are being called and what data is being utilized and potentially leaving and organization's immediate control.
2. Architect for agents: Follow WWT's guidance to build a "data fabric" that supports autonomous AI agents, not just chatbots—including implementing AI governance on agents and the data they use or the information they return.
3. Zero trust for AI: Apply the same zero trust principles to AI models and agents that are applied to human users—verify every prompt, every time.
4. Operationalize AI costs: Use caching and rate-limiting at the edge to prevent "bill shock" from excessive LLM token usage and optimize storage for large data transfers needed for AI training and inference to avoid egress fees.

– Jillian Anderson-Nix, AI Security Strategist & ARMOR Lead

The Cloudflare advantage

Cloudflare AI Security Suite, part of Cloudflare connectivity cloud, offers a broad array of capabilities that support an organization's strategy, security, and governance of evolving AI systems and architectures through a simplified, unified platform experience.

The future of AI in the enterprise is agentic, integrated, and pervasive. By combining WWT's deep architectural expertise to build the high-performance fabric needed for AI along with Cloudflare's global security network, enterprises can stop questioning "How do we stop AI?" and start pondering, "How fast can we go?"