Short answer
Use prompt inspection and DLP. FirstNet sets up Cloudflare to check prompts and uploads in real time and stop sensitive data reaching AI.
In detail: Stop Sensitive Data Going Into AI Tools
Blocking every AI tool rarely works: people find a way around it. The better approach is to let staff use approved tools and stop the data that must not leave.
What the controls do:
- Inspect prompts as they are sent and block or log sensitive content such as ID numbers, financial data, customer information and source code.
- Apply DLP to file uploads as well as typed prompts.
- Detect risky behaviour, such as attempts to jailbreak a model.
- Allow, block, isolate or redirect each AI app per user and group, steering people to the approved tool.
These controls apply to managed devices running the Cloudflare WARP client. For your own AI applications and agents, redaction can also be built into the application itself, so it works on any device. A FirstNet specialist can design the rules with your risk owners.