In its latest round of announcements, Cloudflare is dedicated its latest solutions towards AI, specifically new AI governance tools for IT teams while also open-sourcing Cloudflare OS, an AI workspace designed to help businesses securely deploy AI across their organizations.
The first addition is Identity-Aware AI Gateway that gives organizations visibility into exactly who is using AI services and how they’re being used. Instead of relying on shared API keys, every AI request can now be tied to an authenticated employee or automated agent. IT teams can set spending limits for individuals or departments, monitor AI usage patterns, detect suspicious behavior, and automatically filter sensitive information before it reaches external AI providers.
The accompanying User Insights feature builds on that by learning each user’s normal AI usage over time. If someone suddenly starts generating unusually high AI costs or behaves outside their typical pattern, administrators receive alerts before the situation escalates. It can even identify opportunities to reduce costs by recommending lighter AI models for simpler workloads instead of always using premium models.
On the other hand, open-source AI workspace segment welcomes Cloudflare OS. Initially an internal tool, they are releasing it to the public to allow all users connect AI directly to an organization’s internal systems, documents, and workflows, enabling employees to research information, generate documents using live company data, automate repetitive tasks, and even build internal applications without requiring software development expertise.
Built around the company’s Zero Trust architecture, where every user and AI request must be verified before accessing company resources. AI agents receive only the permissions required for each task, while a system called Gatekeepers lets administrators control exactly what internal systems AI can view or modify.
Cloudflare OS also works alongside AI Gateway to support multiple AI models while giving administrators centralized visibility into usage, budgets, and model selection. This allows businesses to balance performance with operating costs instead of defaulting to the most expensive AI models for every request.
It is available today as an open-source project, while managed deployments can be inquired here.
