Prompts
Available nowInspect supported typed or pasted prompts before submission. Test sensitive and allowed examples separately; a typed prompt test does not establish file-upload coverage.
Generative AI DLP
DLP for generative AI,prompts and files included
See how sensitive prompts and documents can be blocked or redacted before submission. Aona's deployed browser plugin or native app protects supported inputs; coverage depends on the app, client and input path.
SAISIE PROTÉGÉEChamps sensibles remplacés
[CLIENT] et [COMPTE] sont protégés dans cet exemple synthétique.
Exemple synthétique. Validez le client et le chemin de politique concernés.
AI-native
Sensitive-data detection
10,000+
AI tools in the catalogue
7 regions
Aona-managed residency options
SOC 2
Type 2 report available
How it works
From sensitive input to a policy response
Follow the steps on a supported submission path. Existing DLP can also inspect AI traffic; compare the actual action and configured response.
- 1
Inspect the supported input
Aona's deployed browser plugin or native app inspects supported prompts and uploads before submission. The app, client and input path determine what is covered.
- 2
Classify the sensitive content
The configured policy checks for sensitive data such as personal information, financial records, source code and credentials. Use representative synthetic inputs to test the categories you need.
- 3
Enforce your policy in real time
On a supported path, block the submission or redact sensitive entities according to policy. Check what the employee sees and which event is recorded, as well as the protected output.
What's covered
Prompts, files, and supported GenAI tools
Evaluate each input separately. File format, client and upload method matter.
Word & Excel uploads
Available nowReplace sensitive entities in supported DOCX and XLSX uploads while preserving layout. Include tables, cells and document structure in the output review.
PDF uploads
Available nowLayout-preserving PDF redaction is available on supported upload paths. Test your representative PDF structure rather than inferring identical behaviour from a Word file.
Other AI tools
Check input pathApply policies to supported AI clients and input paths. Aona’s catalogue of 10,000+ AI tools adds discovery and risk context; it does not imply uniform enforcement across every listed tool.
A document can carry sensitive data beyond the visible prompt. See how layout-preserving file redaction works.
A synthetic example
Remove the sensitive fields. Keep a useful document.
Switch between the original and protected example. The contact and account are replaced; the project update remains readable.
What to check in a real file test
Review entity replacement, tables, formatting and the output file. Aona supports layout-preserving DOCX, XLSX and PDF redaction on supported upload paths. This illustration is not an installed-product result.
PROJECT SUMMARY
Customer update
- Contact
- [NAME]
- Account
- [ACCOUNT]
The project review is scheduled for Monday.
Beyond blocking
Make the next step clear to employees
The visible response should help an employee understand the policy at the point of use.
Coaching at the moment of risk
Show the reason for the policy response and the guidance available on the supported path. Review it with the employee group before widening the rollout.
Hard blocks where they belong
Where policy requires a hard block, verify that the synthetic submission is stopped. Record the control's scope and review the available event evidence separately from the employee message.
92.8%
Shadow AI prompts fell from 446 to 32 in 30 days after platform blocking and pre-submission guardrails.
Trust
Security evidence, data residency you choose
Review the assurance report and your selected processing configuration. Aona hosting does not determine the AI provider's processing location.
- SOC 2 Type 2 report
- Independently attested controlsSecurity, availability, and confidentiality
- 7 regions
- Aona-managed data residencyAustralia, France, UK, Germany, US, Singapore, Hong Kong
- AI-native
- Sensitive-data detectionPII, financial data and source codeExplore historical Orbital evaluations
Go deeper
Per-tool guides and the AI DLP hub
This page covers DLP for generative AI as a whole. Each major tool has its own deep dive, and the hub organises the full cluster.
AI DLP hub
Choose an app-specific guide or plan the control you need to evaluate.
DLP for ChatGPT
Scope the account, client and sensitive prompt or upload to test.
DLP for Microsoft Copilot
Separate supported prompt controls from Microsoft 365 content permissions.
DLP for Gemini
Check the Google account, Gemini surface and input path.
DLP for Claude
Evaluate Claude inputs alongside the account's native controls.
Switch from Legacy DLP
How to add the AI layer without ripping out the DLP you already run.
FAQ
Generative AI DLP questions
What is generative AI DLP?
Generative AI DLP is data loss prevention that operates where generative AI usage actually happens: the prompt box and the file upload. Aona inspects supported prompts and attachments through a deployed browser plugin or native app, classifies sensitive data such as PII, financial records, source code, and confidential documents, then blocks or redacts it in real time, before it reaches ChatGPT, Microsoft Copilot, Gemini, Claude, or another supported generative AI service. Controls depend on the supported client and input path.
How does DLP for generative AI work?
A deployed browser plugin or native app inspects a supported prompt or upload before submission. Detection checks the input against configured sensitive-data categories. The policy then determines the supported blocking or redaction response and employee guidance. Existing DLP may also inspect AI inputs through endpoint, browser or network controls; compare the exact app, client, input path and policy conditions.
Can generative AI DLP redact files, not just prompts?
Yes. Aona supports layout-preserving redaction for DOCX, XLSX and PDF files on supported upload paths. Test representative synthetic documents and review the protected file's entities, layout and structure. Confirm other file types and actions separately; a successful prompt or DOCX test does not establish their coverage.
Does generative AI DLP block employees from using AI?
The configured policy determines the response on the supported path. It may require a hard block or permit a redacted submission. Test an allowed input as well as a sensitive one, and review the guidance the employee receives. A particular outcome is not guaranteed across every tool or input path.
Where is prompt data processed and stored?
For an Aona-managed regional deployment, Aona offers 7 regions: Australia, France, the UK, Germany, the US, Singapore, and Hong Kong. Backend hosting and prompt-processing placement are separate choices, so confirm the supported configuration for the required feature and integration. Third-party AI-provider processing and residency are separate from Aona's configuration. Aona has a SOC 2 Type 2 report, with evidence available through its Trust Center.
See it in action
Choose a sensitive input.
Define the protection it needs.
Bring the app, client and a synthetic prompt or file. We’ll scope the policy response, employee experience and evidence to check in a guided evaluation.
30-day free trial · SOC 2 Type 2 report