As artificial intelligence becomes embedded in everyday workplace workflows, organizations face a growing security challenge: employees may use AI tools without clear visibility into which platforms they access or what information they submit. The featured 30-minute session demonstrates how Cloudflare Zero Trust and Cloudflare AI Gateway can provide organizations with greater visibility, policy enforcement and control over workplace AI usage, including prompts, AI applications and MCP server access.

The Rise of Shadow AI in the Workplace

The session begins with the problem of “shadow AI,” where employees can access AI applications without centralized organizational controls. This creates visibility gaps around the tools being used and the information being shared with them.

The presentation also highlights prompt exposure and MCP servers as additional areas requiring oversight. Without policies operating at the network and identity layers, organizations may have limited ability to determine which AI services employees are using or establish rules governing their access.

Cloudflare One Brings AI Usage Under Centralized Control

The demonstration presents Cloudflare One as the platform for addressing these challenges. One of the featured capabilities is shadow IT analytics, which allows organizations to see which AI tools their teams are accessing.

The session then demonstrates Gateway rules that can block or redirect unsanctioned AI applications. This provides an enforcement mechanism for organizations that want employees to use approved AI services while restricting access to applications that have not been authorized.

The approach is designed to provide policy controls without requiring organizations to build custom code for every AI application they want to govern.

DLP Adds Protection Against Sensitive Information Exposure

Another major focus is data loss prevention, particularly for interactions with ChatGPT. The demonstration shows how DLP rules can prevent employees from submitting sensitive information such as personally identifiable information or credentials through AI tools.

This capability addresses a growing concern for organizations adopting generative AI: employees may unintentionally include confidential information in prompts. Rather than relying entirely on employee awareness, policy-based controls can provide an additional layer of protection.

AI Gateway and MCP Controls Extend Security to AI Agents

The session also demonstrates Cloudflare AI Gateway as a security and observability layer between AI agents and large language model providers. The presentation highlights capabilities related to jailbreak blocking and cost visibility, allowing organizations to monitor and control AI activity while gaining greater insight into usage.

MCP server access is another area covered in the demonstration. The session shows how organizations can apply Zero Trust controls to govern which MCP servers AI agents are permitted to access, extending existing application-access policies into emerging agent-based workflows.

Market Outlook

As organizations move from experimenting with AI toward broader workplace deployment, controlling access and protecting information will become increasingly important. The session illustrates a model in which AI applications, prompts, agents and MCP servers can be brought under centralized security policies rather than managed independently by employees. For businesses expanding their use of AI, the ability to combine visibility, access controls, DLP and observability could become an increasingly important part of enterprise AI infrastructure and cybersecurity strategies.