What Ward does today, and what is coming.

Ward decides locally, at the moment a user pastes, drops, uploads or submits something, using a signed policy cached on the device. A rule combines who (user, groups), device (managed or not), application (and whether it is sanctioned), account (corporate, personal or unrecognised), action and data (local DLP classification), and returns allow, log, warn, justify or block.

Area status: Available Stops sensitive company data from leaving through AI prompts, pastes, uploads and form submissions — decided in the browser, at the moment of the action.

• AI prompt protection

  AvailableValidated on real Edge

  Inspects text pasted into or submitted to AI tools (Enter, send button, form submit) and applies the policy decision before the prompt is sent.

  Validated on chatgpt.com in real Edge. Apps with unusual send mechanics may not be intercepted on submit; paste and upload interception do not depend on the app.

• Paste and drag-and-drop inspection

  AvailableValidated on real Edge

  Every trusted paste and drop (text and files) is inspected locally. Keystrokes are never observed.

  Paste blocking validated in real Edge on Windows and macOS.

• File upload inspection

  AvailableAutomated tests

  Reads the text of files at upload time: text-like formats, .docx/.xlsx/.pptx, and PDFs (parsed in an isolated extension frame with time and size limits).

  Not inspected: images, legacy Office formats, archives, encrypted or scanned PDFs, files over the size budget (10 MB by default). Rules can choose to block content that could not be inspected.

• Local DLP detectors and classifications

  AvailableAutomated tests

  Built-in detectors for payment cards (Luhn-checked), U.S. Social Security numbers, email addresses, U.S. phone numbers, cloud and SaaS keys and tokens, private keys, database connection strings with credentials, JWTs and generic secrets — plus custom keyword dictionaries and regular expressions. Detectors return counts, never values.

  Payment card detection is part of the real-world validation.

• Corporate vs personal account context

  AvailableAutomated tests

  Reads the account an app displays (for example Google’s account button or a SharePoint tenant host) and classifies it as corporate, personal or unrecognised, so policies can treat a personal account differently from a work one.

  ChatGPT and Claude usually do not display the signed-in email, so they classify as unrecognised. Policies should treat personal and unrecognised alike for those apps.

• App and AI discovery

  AvailableAutomated tests

  Daily roll-up of which applications are used, with which account type, against a built-in catalog of about 65 apps. Unknown hosts that look like AI tools are treated as unreviewed generative AI.

  Records hostnames only. Organisations can restrict discovery to catalog applications.

• Graduated decisions

  AvailableValidated on real Edge

  Allow, log, warn, require a business justification, or block. The strongest matching rule wins; relaxations are explicit, scoped, audited, expiring exceptions.

  Block validated in real Edge; warn and justify are covered by automated browser tests.

Ward inventories every extension on enrolled browsers and tracks what each one is, where it comes from, what it may do and what changed. Blocking is enforced by Microsoft Edge policy, which remains authoritative.

Area status: Available Shows every browser extension in the organisation, where it comes from, what it may do and what changed — and blocks it through Microsoft Edge policy when an administrator decides to.

• Extension inventory

  AvailableAutomated tests

  Name, ID, version, permissions and host access of every extension on enrolled browsers, with affected devices and users.

  Exercised end to end in headless Chromium; not yet against real store extensions updating on real Edge machines.

• Publisher and store history

  AvailableAutomated tests

  Publisher, source (Edge Add-ons, Chrome Web Store, self-hosted, unpacked) and store status over time, including removal from a store.

  Live store lookups are opt-in and rely on an unofficial Edge Add-ons endpoint and the public Chrome Web Store listing page; only the extension ID is sent.

• Permission-change alerts

  AvailableAutomated tests

  Alerts when a new version gains high-risk permissions or all-sites access, changes publisher or update source, or disappears from its store.

• Explainable risk score

  AvailableAutomated tests

  A 0–100 score from declared permissions, site access, source, store status and recent changes; every factor is shown with its points.

  A risk score, not a verdict: Ward does not label an extension “malware” from these signals.

• Administrator blocking via Edge policy

  AvailableAutomated tests

  An administrator marks an extension Blocked; Microsoft Edge enforces it through policy (ExtensionSettings on macOS, ExtensionInstallBlocklist on Windows).

  On Windows the generated Intune script must be redeployed after a change; on macOS the Intune profile is pushed after preview.

• Malware scanning and reputation

  Coming

  Static analysis of extension packages and reputation data.

  Early access on request.

• Automatic blocking

  Coming

  Rules that block extensions automatically, for example on a known-malicious verdict.

  Early access on request.

Area status: Coming Blocking of phishing, malware and scam sites is coming. Today, company policy can already block or warn on navigation to specific applications and categories.

• Policy-based site and app controls

  AvailableAutomated tests

  Block, warn or require a justification when users navigate to specific applications or categories (for example unsanctioned AI tools). Blocks of known hosts are applied before the page is requested.

  Blocks that depend on the account type, and blocks of unknown hosts, are applied in the page shortly after it starts loading.

• Phishing, malware and scam site blocking

  Coming

  Threat-category blocking based on vetted feeds and organisation deny lists.

  Early access on request.

Area status: Coming Download reputation is coming. Today, company policy can already block, warn on or log downloads by source site and file type.

• Download rules by source site and file type

  AvailableAutomated tests

  Policies act on the source application, file name, type and size of a download before it completes.

  Download contents are not inspected.

• Download reputation

  Coming

  Checking download sources against threat data before the file is saved.

  Early access on request.

• Admin console

  AvailableValidated on real Edge

  Devices, discovery, policies (sentence builder with simulator), classifications, alerts, investigations and extension intelligence in one web console.

  Enrolled devices and blocked events were observed in the console during the real-world validation.

• Roles and audit log

  AvailableAutomated tests

  Five administrator roles (Owner, Security Admin, Policy Admin, Analyst, Read Only), enforced by the server. Append-only audit log of administrative actions.

• Signed, versioned policy

  AvailableValidated on real Edge

  Each change compiles an immutable policy version, signed with ECDSA P-256. Devices verify it, can pin the key via MDM, reject rollbacks and keep the last known-good policy offline.

• Alerts and SIEM webhooks

  AvailableAutomated tests

  Rule-based alerts (policy alerts, repeated blocks, heartbeat loss, risky extensions) delivered to SIEM/SOAR through HMAC-signed webhooks.

  No behavioural analytics (UEBA).

• Microsoft Entra ID directory sync

  AvailableAutomated tests

  Users, groups and memberships read from your tenant (read-only Graph permissions) so policies can target groups.

  Tested against mocked Microsoft Graph; not yet validated against a live tenant.

• Administrator single sign-on with Entra ID

  AvailableAutomated tests

  OpenID Connect with PKCE. Administrators must already exist in Ward; SSO never creates them.

  Tested against mocks; not yet validated against a live tenant.

• Device user verification with Entra ID

  AvailableAutomated tests

  Proves which directory user is signed in to the browser using an Entra ID token. Unverified users get every restriction but no identity-scoped exception.

  Tested against mocks; not yet validated against a live tenant or in Edge.

• Microsoft Intune — Windows

  AvailableValidated on real Edge

  Ward generates a PowerShell platform script (and a .reg file) that force-installs the extension and delivers its managed configuration.

  Validated on a real Intune-managed Windows 11 VM: zero-touch enrollment, not removable by the user. Delivery is manual (you upload the script).

• Microsoft Intune — macOS

  AvailableAutomated tests

  Ward generates a .mobileconfig profile you upload as a custom profile. Optionally, after you authorise a dedicated app registration, Ward can create and update the profile via Microsoft Graph — only after you preview and confirm each change.

  Not yet validated through a real Intune tenant; the Graph push is tested against mocks only.

• Microsoft Edge Add-ons store or self-hosted

  AvailableValidated on real Edge

  The extension can be published to Edge Add-ons or self-hosted as a signed package with an update manifest.

  The validation used a self-hosted package with a development signing key; production requires HTTPS and a release key.

• Other MDM / Google Admin console

  Available

  Any tool that can set Edge or Chrome extension policies can deliver the same keys.

  Documented, not validated.