AI data loss prevention

Stop company data leaking into ChatGPT and other AI tools, using the Microsoft 365 security you already pay for.

Staff paste customer records, contracts, code and financial files into public AI tools every day. We find where it happens, then block or warn on sensitive data in AI prompts and uploads with Microsoft Purview, Defender and Intune, without taking AI away from your people.

Book an AI Data Risk Check
Connected systems with a checkpoint that stops sensitive data before it leaves.
The initiative

Move from a difficult technology challenge to a system your teams can use.

For IT heads, security leads and business owners at companies of 20 to 500 staff on Microsoft 365 whose people already use AI at work. We begin with the business context and the systems already in place, then shape a delivery path around what needs to work.

  • AI Data Risk Check
  • Purview DLP for AI prompts and uploads
  • Browser and endpoint controls for AI sites
  • Block personal AI accounts, allow approved AI
  • Sensitivity labels for the data that matters
  • Monthly AI usage and incident report
Explore relevant engineering stories
When this solution fits

Start with the work your team needs to make possible.

AI tools are useful, so people use them with real work: a customer list to clean up, a contract to summarise, code to fix, a board pack to shorten. Much of it happens in personal accounts that your IT team cannot see. Blocking every AI site pushes people to their phones. The workable answer is to detect sensitive data on its way into AI tools, warn or block in the moment, and give staff an approved AI they can use freely.

For IT heads, security leads and business owners at companies of 20 to 500 staff on Microsoft 365 whose people already use AI at work. The first useful step is to define what needs to change, how the current environment shapes the work and which outcome the team needs to be able to assess.

Typical workstreams

What the engineering may involve.

The exact scope depends on your systems and constraints. These workstreams describe common parts of this solution area.

01

AI Data Risk Check

Discover which AI tools are used, with which accounts, and what sensitive data reaches them, then rank the top fixes in a short report.

02

Leak protection setup

Purview DLP for AI prompts and uploads, browser and endpoint controls, sensitivity labels and blocking of personal AI accounts, rolled out in stages.

03

Monitor and improve

Monthly review of AI incidents, new AI apps and policy tuning, with a one-page report for management.

The problem

Your data is already going into AI tools.

Independent security research in 2025 and 2026 shows the same pattern: a large share of AI use at work runs through personal accounts, and the number of incidents where sensitive data is sent to AI apps keeps rising. Most of it is not malicious. It is people trying to get work done faster.

01

Personal AI accounts

Staff sign in to ChatGPT, Gemini and others with personal accounts, so nothing they share is visible to IT or covered by a company agreement.

02

Copy and paste

Customer lists, invoices, contracts, salaries, source code and passwords are pasted into prompts to be cleaned up, summarised or fixed.

03

File uploads

Whole spreadsheets and PDFs are uploaded for analysis, often with more data than the task needed.

04

AI inside other tools

Browser extensions, meeting note-takers and AI features in everyday apps send data to services nobody approved.

05

No record

If a customer, auditor or regulator asks what went into AI tools last quarter, most companies cannot answer.

06

Bans do not work

Blocking every AI site moves usage to phones and home laptops, where there is even less control.

How we stop it

Detect, warn or block in the moment. Keep AI useful.

We use Microsoft Purview, Defender for Cloud Apps, Intune and Entra ID, so there is no new agent to roll out where your licences already cover it. Changes go to a pilot group first.

See the private AI workspace for staff

01

Know your sensitive data

Define what must not leave: customer and personal data, financial records, contracts, source code, credentials. Add sensitivity labels where they help.

02

DLP for AI prompts and uploads

Purview data loss prevention detects sensitive data typed, pasted or uploaded into AI sites, then audits, warns the user or blocks it, by policy.

03

Browser and device controls

Endpoint DLP and supported browser controls on managed Windows and Mac devices, with the Purview browser extension where it is needed.

04

Approved vs unapproved AI

Mark AI apps as sanctioned or unsanctioned in Defender for Cloud Apps; block risky ones and steer people to the tools you approve.

05

Work accounts only

Conditional Access and tenant restrictions so approved AI tools are used with work accounts, not personal ones, where the tool supports it.

06

A safe alternative

Microsoft 365 Copilot or a private AI workspace on Azure OpenAI, so people keep the benefit of AI without the leak.

Licences, honestly

What your Microsoft plan can and cannot do.

AI data protection depends on licensing. We check your plan on the first call and tell you plainly what is possible before any work starts.

01

Business Premium

Device management, Defender for Business, Conditional Access and core DLP. AI site blocking is possible; full prompt-level DLP usually needs a Purview add-on.

02

E3

Strong identity and device controls; prompt-level AI DLP generally needs E5 Compliance or the Purview suite add-on.

03

E5 or E5 Compliance

Endpoint DLP, Defender for Cloud Apps and Purview data security for AI are included.

04

Third-party AI sites

Purview protection for non-Microsoft AI sites such as ChatGPT and Gemini can require pay-as-you-go billing to be enabled. We set it up with a budget.

What you get

Four steps, each one useful on its own.

Start with the check. Most companies see their first real numbers within two weeks.

See what the AI Data Risk Check covers

01

1. AI Data Risk Check

One to two weeks: AI tools in use, accounts, sensitive data shared, Secure Score gaps and the top ten fixes.

02

2. Leak protection setup

Two to four weeks: DLP for AI, browser and device controls, labels and account controls, rolled out in stages.

03

3. Safe AI for staff

Copilot readiness or a private AI workspace, so blocking the risky tools does not mean losing AI.

04

4. Monthly monitoring

New AI apps, incidents, risky users and policy tuning, in a one-page report.

Checklist

Is data leaking into AI tools at your company?

If you cannot answer yes to most of these, it probably is.

Request an assessment
A clear path from scope to operation

Keep decisions connected through delivery.

Move from a defined initiative to an operating system through focused, reviewable stages.

01

Understand the workflow

Agree on the users, business context, systems and constraints that define the initiative.

02

Shape the technical path

Connect architecture, integrations, data, security and delivery decisions to the needs already identified.

03

Build in useful increments

Deliver and review the system in stages so the work can respond to product and operational feedback.

04

Prepare to operate

Consider ownership, monitoring, support and the next improvements as part of the delivery plan.

Discuss this solution
FAQ

Questions about ai data leak prevention

What is AI data loss prevention?

AI data loss prevention (AI DLP) detects sensitive information on its way into AI tools, such as text pasted into a ChatGPT prompt or a file uploaded to an AI assistant, and audits, warns or blocks it according to your policy.

Can you stop employees pasting sensitive data into ChatGPT?

Yes, on managed devices with the right Microsoft licence. Purview DLP can detect sensitive data in prompts and uploads to supported AI sites and warn the user or block the action, while normal AI use carries on.

Do we have to block ChatGPT completely?

No. Most companies allow AI and block only sensitive data, personal accounts or high-risk apps. Blanket bans tend to push usage to personal devices.

Which Microsoft 365 licence do we need?

It depends on the depth of control. Business Premium covers device management and site blocking; prompt-level DLP for AI usually needs E5, E5 Compliance or a Purview add-on. We confirm what your plan can do before any work starts.

Does this work on personal phones and home laptops?

Controls apply to managed devices and work accounts. For personal devices we restrict access to company data through protected apps, so it cannot be copied into personal AI tools.

Will staff see a warning or just be blocked?

You choose per policy: audit only, a warning the user can override with a reason, or a hard block. Most rollouts start with audit and warnings.

How long does setup take?

The AI Data Risk Check takes one to two weeks. The protection setup usually takes two to four weeks with a staged rollout.

Can you work with our existing IT provider or MSP?

Yes. We can deliver alongside your IT team or MSP, or white-label the work for an MSP.

Connected capabilities

The disciplines that make this work.

Explore the specific engineering capabilities that support this solution.