AI Demo Cloudflare AI security demo

Demo scripts

Two demos. Five short scripts that each prove one control, and one continuous walkthrough that follows a single person until the whole picture assembles itself.

Which one to run

Five minutes, or a specific question to answer: pick from the platform controls below. Each is one prompt and stands alone.

Twenty minutes, and an audience you want to convince: run the walkthrough. It is the one people remember, because the person in it never does anything wrong.

1. Platform controls

One prompt each, in this order. Between them they cover every control in the stack rather than the same policy five times over.

Script Cloudflare control What it proves to the customer
Asking the company assistant to help commit a crime unsafe use of company AI AI Gateway Guardrails (violent and non-violent crime, weapons, self-harm) Governs what your staff can make the company's AI do, consistently across every model - instead of relying on each vendor's own refusals. Blocked prompts cost no tokens.
Pasting a colleague's details into a chat accidental exposure AI Gateway DLP (request and response) Covers the everyday incident with no agent in it: someone pasting personal data into a chat. Same profiles as the MCP path, second choke point.
The CEO's home address WorkWeek (HR) · intentional misuse Gateway DLP on MCP traffic (Employee PII) Your apps can be carefully built and still hand an agent a home address. One profile, applied at the network, stops it without touching the app.
The poisoned style guide Nexus (Wiki) · indirect prompt injection AI Gateway Guardrails (prompt injection) + Gateway DLP One insider, or one compromised account, can leave instructions that steer everybody else's AI. Two independent layers stop it, neither relying on the model's judgement.
The application she cannot see at all Ledger (Finance) · access control Cloudflare Access policy + MCP portal scoping The cheapest control of all: the app and its tools are not yours to see. An agent cannot call a tool that was never listed for your identity.

Controls covered by these five:

2. The walkthrough

One person, one afternoon →

Eleven steps, alternating between the app and the agent. Delta reads a rumour about an acquisition and a restructure, looks for the truth in four applications that each correctly tell her nothing, and asks her assistant the same questions instead. It begins with her pasting a colleague's customer list into a public AI to tidy it up, and ends with "What is Project Ironwood?" being answered in a paragraph.

Covers AI Gateway DLP, three Gateway DLP policies on MCP traffic, and Access twice — both as a locked door and as an invisible tool list.

Archive

Nine earlier scripts, kept for reference. Most are variations on Gateway DLP over MCP traffic and are superseded by the walkthrough, which shows the same thing in sequence. Browse the archive →

Before you start