Content Guardrails
Content Guardrails: what your staff send to AI tools, and what should never have left
Staff paste things into AI tools that should never leave the business: a customer’s card number to draft a refund email, a spreadsheet of ID numbers to summarize. Usually nobody finds out.
With Cascadia you get
Reviewed by a person each week
Scanned, not sampled
Blocked or flagged, your choice
Set up and watched by us
No agent on your machines
Your data stays yours
Why teams start here
Reviewed by a person each week
Flagged items are read by us each week rather than counted by a dashboard. Patterns get raised with you, repeat offenders are surfaced quietly, and false positives are tuned out as you learn what your people actually need to paste.
Scanned, not sampled
Scanning happens at the gateway, so it covers every request from every application rather than depending on what is installed on each laptop.
Blocked or flagged, your choice
You choose whether a match is blocked outright or allowed and flagged. Blocking is safer; flagging is easier to live with while people adjust.
What you get
Everything Content Guardrails covers
All of these come with Content Guardrails, set up and looked after by our team.
What is caught
- Card numbers and bank details
- ID numbers
- Financial information in prompts and replies
- Anything your own profiles define
- Both what is sent and what comes back
How it works
- Scanning runs at the gateway, not the device
- Every request, not a sample
- Blocked outright, or flagged and allowed
- Applied per application
- No change to how your staff work
Reviewing
- Flagged items read by us each week
- Patterns raised with you, not just counts
- Repeat offenders surfaced quietly
- False positives tuned out
- Rules adjusted as you learn
Reporting
- A monthly report of what was caught
- Which application it came from
- What was blocked and what was allowed
- Patterns raised with you, not just counts
- Evidence if you ever need it
What sets it apart
- Reviewed by a person each week
- Scanned, not sampled
- Blocked or flagged, your choice
- Set up and watched by us
- No agent on your machines
- Your data stays yours
Want to know what is being pasted into AI tools?
Tell us what you sell and we will start with the questions your buyers ask.
Get startedSee the comparisonsWhat a policy document catches, and what it leaves out
A policy tells your staff what not to paste into an AI tool. It cannot tell you whether they did. The useful part is further in: what actually went out, which application it came from, and what was stopped before it left.
| Trusting people to be careful | Content Guardrails | |
|---|---|---|
| Card numbers | Pasted and forgotten | Card numbers and bank details caught at the gateway |
| ID numbers | No idea either way | ID numbers caught the same way |
| Coverage | Depends on the laptop | Every request scanned, not a sample, with no agent on your machines |
| Evidence | None | A monthly report of what was caught and which application it came from |
| Review | Nobody reads it | Flagged items read by us each week, false positives tuned out |
| Upkeep | A policy document | Rules adjusted as you learn, repeat offenders surfaced quietly |
What clients say about working with Cascadia
Want to know what is being pasted into AI tools?
Talk to us
Who Content Guardrails fits best
Teams that cannot rely on what is installed on each laptop
Scanning happens at the gateway, so it covers every request from every application rather than depending on what is installed on each machine. You choose whether a match is blocked outright or allowed and flagged, and a person reviews the matches each week.
Companies unsure what staff paste into AI
Staff paste things into AI tools that should never leave the business, and usually nobody finds out that it happened.
Agencies reporting this for their clients
You can sell the outcome without hiring for it. We run the technical and content work under your brand, deliver reporting you can pass straight to a client, and stay out of the relationship.
How it works
1. We switch scanning on
We start with the predefined profiles for financial and identifier data, then tune them to what your business actually handles.
2. We tune it to your business
Scanning is switched on at the gateway, tuned to the data your business actually handles, and set to block or to flag according to what you choose. Your staff carry on using the same tools.
3. You get the picture every month
Each week we read what was flagged, and each month you get a report of what was caught and where it came from.
Ask us
Answers to common Content Guardrails questions
Weighing us against another option? Our comparisons take the main ones in turn.
What is Content Guardrails?
Scanning that sits inside your AI Gateway and reads every prompt and every reply for card numbers, bank details and ID numbers, and anything else your own profiles define. What should not leave the business is blocked outright or allowed and flagged, before it goes out rather than after.
How is this different from a policy telling staff to be careful?
A policy relies on everyone remembering it, and nobody finds out when somebody forgets. Scanning at the gateway checks every request instead, so a card number pasted in to draft a refund email is caught rather than trusted not to happen.
Does this need AI Gateway to work?
Yes, it depends on AI Gateway, because the scanning happens there. On its own it has nothing to scan.
What do you install on our side?
It needs AI Gateway in place. Beyond that, nothing is installed and your staff carry on using the tools they already use.
How long before I see results?
Setup starts within two business days, and the first full picture arrives within a week.
Do you keep a record of what was caught?
Yes. What was caught, when, and from which application, kept for 90 days. The content itself is held only as long as you want it to be.
Why scan at the gateway instead of on each device?
Scanning at the gateway covers every request from every application, rather than depending on what happens to be installed on each laptop.
Does this slow anything down for staff?
Your staff carry on using the tools they already use and nothing is installed on their machines, so there is nothing sitting in the way. The only time anyone notices is when a match is blocked, and whether a match is blocked or simply flagged is your choice.
What happens to an item that gets flagged?
We read the flagged items each week rather than leaving them in a queue. Patterns get raised with you instead of a bare count, repeat offenders are surfaced quietly, and false positives are tuned out so the profiles get more accurate rather than noisier.
What do you need from me to get started?
AI Gateway in place, and a conversation about what counts as sensitive in your business.
Can we decide what counts as sensitive?
Yes. We start with the predefined profiles for financial and identifier data, then tune them to the data your business actually handles. Anything your own profiles define is caught as well, and the rules are adjusted as you learn what matters.
Is this worth it for a small team?
A small team is usually the case for it rather than against it. There is nobody whose job it is to watch what gets pasted where, and one card number in a prompt is the same problem at any size. Because the scanning runs at the gateway, there is nothing to roll out and nothing for anyone to administer.
Who sees what was caught?
You do. The monthly report names what was caught, which application it came from, and what was blocked as against what was allowed. For most businesses it is the first honest picture of how their AI tools are being used, and it is evidence if you ever need it. The content itself is held only as long as you want it to be.
Do I need managed WordPress hosting or maintenance to buy this?
No. It runs as a standalone monthly service on whatever hosting you already use. It does need AI Gateway in place, because the scanning happens there. It is also part of AI Pro, if you would rather have the AI services on one invoice.
Ask Us Anything
We’d love to hear from you!