Locke vs Microsoft Purview Information Protection

When Microsoft Purview Information Protection fits

  • Your organisation is deeply invested in Microsoft 365 and needs sensitivity labelling, encryption, and rights management across email, Teams, SharePoint, and Office documents.
  • You need centralised policy management, compliance manager dashboards, and eDiscovery that spans the full Microsoft 365 estate.
  • Your primary AI governance concern is Microsoft Copilot. Purview's AI Hub provides prompt auditing and policy enforcement within the M365 boundary.

When Locke fits

  • Employees use browser-based AI tools outside the Microsoft 365 boundary, such as ChatGPT, Claude, Gemini, Grok and Perplexity, where Purview has no reach.
  • You need a control that acts at the moment of composition, before text is submitted, rather than after the document is classified.
  • You want a zero-knowledge control: no prompt content is logged by the privacy vendor, unlike Purview's audit logs which capture prompt text.

Side-by-side

DimensionLockeMicrosoft Purview
Coverage of non-Microsoft AI toolsWorks on supported browser-based AI chats: ChatGPT, Claude, Gemini and others.Limited to Microsoft 365 apps and Copilot; no coverage of competitor AI tools.
Where controls operateOn the device, inside the browser, at the moment of composition.At the Microsoft 365 layer: documents, email, Teams, SharePoint.
Sensitive-data visibility to the vendorZero-knowledge: Sonomos never sees prompt content.Purview audit logs capture prompt text and policy matches, visible to your admin.
Rollout effortPer-user, on-device Locke Early Access install: days to deploy at scale.Deep integration with M365 tenant configuration: weeks to months.
Best forPreventing sensitive data from reaching any browser-based AI tool.Classifying, labelling, and protecting documents and communications within M365.

Bottom line

Purview and Locke protect different surfaces and work well together. Purview governs the Microsoft 365 estate: documents, email, Teams and Copilot. Locke screens the moment someone uses another AI tool it covers, such as pasting a labelled Word document into ChatGPT.

Further reading