30 jours d'essai gratuit, risques IA générative :Commencer
Aller au contenu principal

AI SECURITY COVERAGE

AI security coverage
Know where you stand.

See where Aona discovers AI use, applies prompt policies and protects files. Find the right coverage for your employees’ browsers, desktop apps and AI agents.

For security and IT teams planning a workforce AI rollout.

Choose your employee surface

Chrome, Edge, Firefox & Safari

Inside the browser

Production plugin

Discover AI use

On devices with the plugin installed

Endpoint required

Apply prompt policies

On supported AI submission paths

Route-specific

Redact sensitive files

DOCX, XLSX and PDF upload paths

Validate output

Start with ChatGPT, Claude, Gemini, Copilot or Bing. Confirm the exact interface, action and policy, not just the tool name.

Coverage guide · not a live device scan

FOUR DIFFERENT QUESTIONS

A known tool isn’t
a protected workflow.

Separate catalogue recognition, observed use and tested policy outcomes. Each needs its own evidence.

  1. 01

    Known

    The tool is in the catalogue.

    Not proof of employee use.
  2. 02

    Observed

    Activity appears on a deployed endpoint path.

    Not proof of a policy action.
  3. 03

    Evaluated

    A configured policy evaluates a prompt or file.

    Validate the exact route.
  4. 04

    Verified

    A current test documents the outcome.

    Keep its scope and evidence date.

THE COVERAGE REFERENCE

Find your path.
Check the details.

Start with the capability you need. Then validate the app, endpoint and action that make it work.

AI tool discovery

10,000+ AI tools tracked

Catalogue breadth
Confirm observed use through the installed browser plugin or native app. No endpoint means no Shadow AI visibility.

Browser prompt policies

Chrome · Edge · Firefox · Safari

Supported routes
Active governance covers top-tier ChatGPT, Claude, Gemini, Copilot and Bing. Test the exact interface, policy and submission action.

File redaction

DOCX · XLSX · PDF

Documented formats
Check the upload route, sensitive-data replacements and resulting document layout. Format support is not every possible file.

Native AI applications

Windows & macOS

Release-specific
IT teams can deploy the Aona endpoint app on Windows and macOS with their preferred software deployment tools. Intune is one option; direct installation is also available on both platforms. Confirm the provider, OS, input path and installed release for the controls you need.

AI agent inspection

Native endpoint app

Limited rollout
Confirm runtime and privacy scope. Inspection does not imply remote agent blocking, fleet-wide scanning or agentless API coverage.

Unverified stays unverified. Record the release, configuration and evidence date before expanding a rollout.

YOUR EVALUATION, MADE PRACTICAL

Start with a test.
Leave with evidence.

Use a synthetic example, agree the result you need, then record what actually happened.

ChatGPT web · managed browserSynthetic input

Protect the customer detail, keep the request useful.

Summarise the renewal notes for Morgan Hale. Contact: morgan.hale@example.com. Keep the next steps concise.
Download prompt fixtures

WHAT TO CHECK

  1. 01

    Select the person-name and email rules for the test group.

  2. 02

    Confirm the configured response before the request is sent.

  3. 03

    Compare the employee notice with the recorded policy outcome.

Record the browser, extension version and ChatGPT interface. Repeat typed, pasted and attached inputs separately.

Get the acceptance checklist
Client contract checks 2026-09-19

These checks run real client functions with mocked dependencies. They do not prove detection accuracy, upload coverage, an installed release or end-to-end Sentinel ingestion.

Policy action handling

Block priority, redaction responses and silent flags with supplied policy responses.

12/12 checks passed

ChatGPT document request

DOCX restoration request uses the stateless replacement contract.

1/1 checks passed

Claude document request

DOCX restoration request uses the stateless replacement contract.

1/1 checks passed
Download the scoped check record

MAKE COVERAGE TANGIBLE

Bring one workflow.
Leave with a clearer decision.

A useful coverage review follows the work your people actually do.

  1. 01

    Name the path

    AI tool, browser or app, OS, and intended action.

  2. 02

    Test with synthetic data

    Run the prompt or upload against the chosen policy.

  3. 03

    Record the decision

    Review the outcome, evidence and unresolved gaps.

YOUR COVERAGE REVIEW

One path. A complete picture.

Application
ChatGPT
Employee surface
Chrome + Aona plugin
Input to test
DOCX with synthetic identifiers
Policy action
Redact before upload
Inspect the redacted outputReview recorded policy evidenceDocument any unsupported behavior

Illustrative test plan, not a completed test or coverage guarantee.

COVERAGE MEETS DEPLOYMENT

Where it runs
is a separate choice.

Choose backend hosting and prompt processing independently. Verify that your required features support the selected combination.

01 / BACKEND HOSTING

Customer cloud / On-premises / Aona-managed

02 / PROMPT PROCESSING

User device (edge) / Customer cloud or on-premises / Aona-managed

Aona’s configuration does not determine where a third-party AI provider processes data.

BEFORE YOUR PILOT

The questions
worth asking.

Does a catalogued AI tool mean it is discovered or controlled?

Aona tracks 10,000+ AI tools in its catalogue. Catalogue recognition is not observed employee use or active enforcement on your devices; those require an installed client and a supported application and input path.

What must be deployed to observe employee AI use?

The applicable browser plugin or native endpoint client must be installed and configured on the managed device. Aona does not provide an agentless network, identity-log, or SaaS-integration scan for this purpose.

Which browser and prompt-policy paths are supported?

The Aona browser plugin supports Chrome, Edge, Firefox and Safari. Confirm the intended browser, application and input path; browser support does not establish feature parity. Active prompt policy enforcement is focused on top-tier ChatGPT, Claude, Gemini, Copilot and Bing paths; confirm the exact interface, action, data category and current release before rollout.

Which file formats can be evaluated?

DOCX, XLSX, and PDF are documented layout-preserving redaction formats. Validate the intended upload route, policy, and resulting file with a synthetic test.

What are the native and agent limits?

Native provider, transport, OS and release coverage varies. IT teams can deploy the Aona endpoint app on Windows and macOS with their preferred software deployment tools. Intune is one option; direct installation is also available on both platforms. AI-agent inspection is native-endpoint only and in limited rollout; it is distinct from agent-action control.

Does backend hosting determine prompt processing?

No. Backend hosting and prompt processing are independent choices. Confirm the supported feature and integration combination for the selected configuration, and assess third-party AI-provider processing separately.

START WITH YOUR ENVIRONMENT

Map coverage for your environment

Bring the app, surface, endpoint, policy action, and file type you need to evaluate. We will map the current evidence and name what remains unverified.

Map coverage for your environment
AI Security Coverage: Discovery, Prompts and Files | Aona