A New Security Model for Federal AI

Artificial intelligence (AI) has moved quickly from federal experimentation into daily operations. Employees are using it to research and analyze information, developers are embedding it in applications, and agents are beginning to interact with data, APIs, and mission systems on a user’s behalf. These use cases and access routes do not carry equal consequences, yet many agencies still rely on fragmented controls and approval processes designed for systems that change more slowly and act more predictably.

Explore how federal leaders can evolve their security operating models to build AI without losing control. Anish Patel, head of federal at Cloudflare, will discuss how agencies can advance AI at the speed of its impact and consequences, extend visibility across human and machine activity, and apply controls that adapt as permissions, behaviors, and use cases change. Across a video, podcast, and written Q&A, the program will examine how agencies can move beyond point-in-time approvals and fragmented safeguards toward a more continuous, unified approach that supports AI adoption while protecting public trust, service continuity, and mission outcomes.

WATCH:
Setting the Boundaries for Federal AI

AI agents introduce a new Zero Trust principal – one that can act with delegated authority at machine speed and be influenced by the information it consumes. Watch the video to learn how federal agencies can apply visibility, least-privilege access, and human-approved policy enforcement to keep AI activity bounded, observable, and reversible.

LISTEN:
Governing AI in Real Time

As AI becomes part of federal workflows, applications, and mission systems, agencies need a security operating model that can adjust as quickly as the technology and its uses evolve. Models change, data connections expand, and agents may perform thousands of actions in the time it takes a person to complete one. Agencies therefore need to complement formal authorization with continuous evidence and live enforcement.

In this podcast, Anish Patel, head of federal at Cloudflare, will discuss how agencies can progress from point-in-time approvals toward continuous, risk-informed AI security. The discussion will explore how agencies can adjust oversight as AI systems, permissions, and risks change, and how security, technology, data, and mission teams can support the new security operating model together. Patel will also discuss how consistent visibility and policy enforcement can help agencies accelerate appropriate uses of AI while strengthening accountability and resilience.

READ:
Securing Federal AI at the Speed of Change

Federal agencies are moving AI capabilities from controlled pilots into environments where they interact with sensitive data, production applications, external users, and mission systems. Federal agencies already have strong foundations in Zero Trust, the NIST Risk Management Framework, continuous monitoring, and cloud authorization. The task is to extend those practices to systems whose models, data connections, permissions, and behavior may change after deployment.

In this written Q&A, Anish Patel, head of federal at Cloudflare, will offer practical guidance for making AI activity bounded, observable, and reversible. The discussion will cover initial access decisions, cross-environment policy enforcement, change management, ongoing authorization evidence, and concrete actions federal IT and program leaders can take now.

Get Your Copy

Please provide the information below. All fields are required.