Content Guardrails
A customer’s card number goes into a prompt to draft a refund email. A spreadsheet of ID numbers goes in for a summary. In most businesses nobody ever learns it happened, and Content Guardrails exists to change that.
With Cascadia you get
With your permission, we read the flagged items ourselves instead of leaving a dashboard to count them. Patterns are raised with you, repeat patterns are surfaced quietly, and settings are tuned as it becomes clear what your people need to send.
Scanning sits at the gateway, so every request your applications send through it is checked, whatever is or is not installed on individual laptops. Text pasted into an AI tool in a browser does not pass through the gateway, so it is not scanned.
A match can be stopped outright or allowed through with a flag. Blocking is the safer setting. Flagging is gentler while people adjust to it.
What you get
Each item below comes with Content Guardrails and is set up and looked after by our team.
Tell us which applications call AI providers, and scanning of each request at the gateway can begin.
A policy tells staff what to keep out of an AI tool, but it cannot show whether they did. What matters is what actually went out, which application sent it, and what was stopped on the way.
| What to compare | Trusting people to be careful | Content Guardrails |
|---|---|---|
| Card Numbers | Pasted, then forgotten | Card numbers and bank details caught at the gateway |
| ID Numbers | No way to tell either way | ID numbers caught the same way |
| Coverage | Varies with each laptop | Every request through the gateway scanned, not a sample, with no agent on your machines |
| Evidence | Nothing to show | A monthly report of what was caught and which application it came from |
| Review | Nobody looks | Flagged items read by us each week, false positives tuned out |
| Upkeep | A policy document that ages | Settings adjusted as you learn, repeat patterns surfaced quietly |
“I’ve always dreaded website management, but Cascadia has done an incredible job with my WordPress site, making it one less thing for me to worry about.”
“I’ve worked with Cascadia for several years now. They are always ready to help in any way I ask and can implement my ideas with ease. A company that values their clients!”
“Cascadia has been great to work with! We recently needed some updates, and Cascadia was quick to get them completed! We highly recommend Cascadia Web Services.”
“Cascadia is very responsive and we’re happy with them as our primary IT vendor.”
“They do great work, been using for years. Prompt responses to requests.”
Because scanning happens at the gateway, coverage does not depend on what each machine has installed. Matches are blocked or flagged as you decide, and a person goes over them every week.
Get StartedMaterial that should never leave the business ends up in prompts, and usually no one finds out. This gives you a record of what went through your gateway.
Get StartedThe monthly report lists what was caught, which application it came from, and what was blocked versus allowed. It is a document you can pass straight to a client.
Get StartedReading the flagged items is our job, so it does not have to be yours.
Scanning begins at your gateway with Cloudflare’s two free profiles, for financial and identifier data. Your staff keep using the same tools.
Get StartedWe set matches to block or flag, as you choose, decide with you whether replies are scanned too, and adjust as the weekly reviews show what your business needs to send.
Get StartedWith your permission, we read what was flagged each week. Every month you receive a report of what was caught and where it came from.
Get StartedPricing
Content Guardrails
What We Do for You
Requires AI Gateway.
Ask us
If you are weighing us against another provider, our comparison pages cover the main ones.
Compare with dope.security, Harmonic Security, and MoreCompare with dope.security and MoreCompare with dope.security and MoreStill have a question?
It is scanning built into your AI Gateway. Every prompt, and if you choose every reply, is checked with Cloudflare’s two free profiles for card numbers, bank details, ID numbers, and other financial information. A match is blocked outright or allowed and flagged, and this happens before the content goes out, not after.
A policy depends on everyone remembering it, and nobody learns when someone forgets. Scanning at the gateway checks each request that goes through it, so a card number in a prompt your applications send gets caught instead of assumed not to happen. Text pasted into a chat tool in a browser does not go through the gateway.
Yes. The scanning runs inside the gateway, so without it there is nothing to scan. It is also included in AI Pro, which bundles the gateway with it.
Nothing, beyond having AI Gateway in place. Your staff keep using the tools they already use.
A gateway covers every request your applications send through it, whatever is installed on each laptop. It does not see a person typing into an AI tool in a browser; if that is your main worry, a browser tool is the better fit, and our comparisons say so.
What your applications send through the gateway, and, if you choose, what comes back. Cloudflare’s two free profiles look for card numbers, bank details, ID numbers, and other financial information. Scanning replies makes streamed replies start a little later.
Within Cloudflare’s two free profiles, for financial and identifier data. We do not add custom profiles, which would need a paid Cloudflare plan. You decide whether matches are blocked or flagged and whether replies are scanned, and we adjust with you as you learn what matters.
That is up to you. Blocking is safer. Flagging is easier while people adjust, and the weekly review shows what is worth tightening.
Staff keep using their existing tools and nothing is installed on their machines. The only time someone notices is when a match is blocked, and whether matches are blocked or merely flagged is your decision.
With your permission, given and taken back on your AI Services page, we read flagged items weekly instead of letting a queue build, and each time we open them it is listed for you. You hear about patterns rather than a bare count, and settings are tuned so the results grow more accurate, not noisier. Without your permission, the review covers counts and patterns only.
You do. The monthly report names what was caught, which application it came from, and what was blocked as opposed to allowed. It often gives a business its first clear look at how its AI tools are used, and it serves as evidence if you need it.
Yes: what was caught, when, and from which application, counted for the report. The content itself stays in your own Cloudflare account and storage for 90 days, locked with a key only you hold.
Setup starts within two business days, and the first full picture arrives within a week.
Often more so. A small team rarely has anyone whose job is watching what gets pasted where, and one card number in a prompt is the same problem at any size. Scanning runs at the gateway, so there is nothing to roll out and nothing for anyone to administer.
No. It runs as a standalone monthly service alongside whatever hosting you use. It does need AI Gateway, because that is where the scanning happens.
We’d love to hear from you!
