AI governance & complianceYour staff use AI. Sensitive data stays put.

RedactWall stops sensitive data leaving through AI tools and email, finds it in the files already sitting in your cloud drives, and turns every decision into evidence your own team can check.

Request a pilot See what it covers

Support agentCustomer service

customer list, pasted into the message

Read here, before it moves

Personal webmailOutside the organization BLOCKED

One refused message, end to end. An illustration, drawn the way your reviewer sees it.

Your controls were built for a world without AI

Staff adopted AI tools faster than anyone approved them.

A customer file pasted into a chatbot leaves as ordinary web traffic.
Blocking yesterday’s list of AI sites misses the shadow-AI tool your staff opened this morning.
When an auditor or examiner asks what left, a log you control is not the same as evidence.

Every path out, and what stands on it

Two paths are decided before anything moves. One finds and reports what is already in your drives. Two more are decided on the device. We publish which is which.

The path out What it does

AI tools and assistants Decides before data leaves Every request and every answer decided before it moves.

Outbound email Decides before data leaves Decided before the next hop accepts the message.

Microsoft 365 and Google Drive Finds and reports Found and reported once the file is already there.

AI agent tool calls On the device A tool call is authorized before it runs.

Laptops and browser On the device Device decisions are re-checked by the server.

A solid marker sits on the path: nothing moves until it has been decided. A dashed marker sits beside the path: the call is decided on the device, or the file is found and reported after it lands. Read the coverage table →

What that looks like in practice

An employee pastes a customer file into a chatbot → stopped before it reaches the model, and the attempt is on the record.
An AI assistant repeats an account number back in its answer → caught on the way back, before staff see it.
Someone emails a customer list to a personal account → refused before the message is accepted for delivery.
An auditor asks what left the organization last quarter → hand over a sealed package covering exactly that period.

Nothing for your developers to learn

One setting per application: the address it already sends to. There is no SDK to adopt.

Your applications Keep the client libraries they use today

One setting

RedactWall Every request and every answer is checked here

AI providers Anthropic, Gemini, Bedrock, and anything OpenAI-compatible

One policy covers every provider your staff reach for. All integrations · How far each provider is proven →

Proof you do not have to trust us for

A decision your reviewer can read, on a record your own team can confirm is intact.

Email to personal account BLOCKED

A loan officer tried to forward a member list to a personal webmail address.

Destination: Personal webmail
Found: 42 member records, 42 account numbers
Policy: Member data to personal accounts
Decided: Before the message was accepted

Recorded on your evidence stream. How your team checks it →

What your reviewer sees. An illustration of a single decision. Built deepest for credit unions →

Your data never enters the record

The record names the kind of sensitive data found. It never holds the values, the message, or the file.

What was sent

What the record holds

Member identifiers found Account numbers found The policy that applied Who decided, and when Not the values themselves Not the message or the file

Illustration. The bars stand for sensitive values found in a message the record never keeps.

Questions we get

More answers, and everything about where the product stands today, on the FAQ page.

Do we have to stop staff using AI?

No. That is the point. Work carries on, and the things that should not leave are stopped. You are not choosing between productivity and control.

Does anything get installed on staff laptops?

Not for the main controls, which run in the cloud. There are optional device and browser components for what leaves from a laptop and never passes a server. They sit on top of the cloud controls, not in place of them.

Do you store our data?

No. Nothing we inspect is kept. The record says what kind of sensitive data was found and what was decided about it, never the data itself.

Will this help with AI compliance and examinations?

It gives you material to prepare with: a sealed package covering a period you name, and a way for your team to confirm the record is intact.

Is this data loss prevention?

It is data loss prevention for the paths AI uses: stopped before it moves in AI tools and outbound email, found and reported after the fact in cloud drives. What the platform covers →

Are you certified?

Not yet. See where each attestation stands today. Disclosures →

How do we know it covers what you say it covers?

We publish what every surface covers, surface by surface. Read the coverage table →

Stop the leak. Not the work.

The engineering team onboards your pilot directly, starting with the surface you care about most.

Request a pilot See what each surface covers

One setting change per application
No SDK for your developers to adopt
Your sensitive data is never stored
Coverage published surface by surface