AI Governance & Security Vendor Wiki

7 min read

Glean

Primary category: entitlement-aware-rag. Secondary: enterprise-ai-assistant.

One-liner — Permission-aware enterprise search and a “Work AI” assistant/agent platform that connects to all your SaaS apps and answers questions over company data — while enforcing each source system’s access controls so a user never sees a document they couldn’t already open themselves.

What it does

Glean indexes content across an enterprise’s apps (Google Workspace, Microsoft 365/SharePoint, Slack, Jira, Confluence, Salesforce, Zendesk, GitHub, etc.) and builds a unified search index plus a “Enterprise Knowledge Graph” of people, teams, documents, and relationships. On top of that sits Glean Assistant — a chat/RAG interface that answers natural-language questions, summarizes, and drafts using retrieved company content as grounding, plus an agent/workflow builder (“Work AI”) for multi-step tasks. The core value is that employees get a single, current, trustworthy answer over fragmented internal knowledge instead of hunting across a dozen tools.

The non-negotiable control underneath all of it: document-level permission enforcement at query time. When Glean ingests content through a connector, it also ingests that source system’s access-control lists (ACLs), group memberships, and org hierarchy, and keeps them synced. At query time it filters retrieval to only the documents the asking user is authorized to see — and, per Glean’s docs, it defers to and never overrides the source system’s permissions (“All document- and app-level permissions continue to be enforced by the original source systems”). This is what makes enterprise RAG safe: without it, an LLM grounded on the corporate index becomes an oversharing machine that surfaces HR files, M&A docs, or comp data to anyone who asks.

Where it sits in the stack

  • Layer: retrieval. Primary category entitlement-aware-rag; also enterprise-ai-assistant.
  • Risk role: controls access to sensitive data. Glean’s job is to make sure the model only ever retrieves data the user is entitled to, so access to private data is scoped to per-user authorization rather than the whole corpus. It does not by itself screen untrusted input or control outbound data flows — those need a runtime/guardrail layer (ai-runtime-security) and DLP/egress controls.
  • Trust zone: lives in the green/internal zone but reaches across every connected SaaS app; the permission-mirroring is precisely what keeps cross-app retrieval from collapsing trust boundaries.

Deployment & architecture

  • Delivery: SaaS (Glean-hosted cloud). Connector-based ingestion with crawlers/APIs into ~100+ enterprise apps; permission/ACL sync via delta updates from source systems and identity providers.
  • Identity: reads groups and members from the IdP (Azure AD / Entra, Google Groups documented; Okta group-based role assignment noted as not yet supported in the RBAC FAQ as fetched). Glean reads identity for group resolution but relies on connector ACLs for document authorization.
  • Interfaces: web app, browser extension, Slack/Teams surfaces, and an API/SDK + Model Context Protocol (MCP) and agent tooling for building on top of the index.
  • Open item: whether a customer-managed VPC / self-hosted / on-prem deployment is offered was not confirmable from the primary docs fetched — Glean is primarily a managed cloud service. Flagged below.

Positioning & differentiators

  • Permission-mirroring is the headline. Glean built its reputation on getting enterprise-search ACL enforcement right at scale before “RAG” was a buzzword; the permissions story is its core moat versus point LLM tools.
  • Horizontal Work AI platform, not a single-app copilot — it spans all connected systems rather than being tied to one vendor’s suite.
  • Nearest neighbors: microsoft-365-copilot (bundled with M365, leans on Microsoft Graph/microsoft-graph permissions but weaker on non-Microsoft sources), gemini-enterprise (Google’s Agentspace, strong on Workspace), and microsoft-graph as the underlying permissioned-retrieval substrate for Copilot. Glean’s pitch against the platform copilots is breadth of connectors and being suite-neutral; the copilots’ pitch against Glean is “already in your license, no extra vendor.” Knostic and others have published research probing whether any of these (Glean included) fully prevent oversharing when source-system permissions are themselves sloppy — the control is only as good as the upstream ACLs.

Ownership, funding & M&A

  • Independent, venture-backed. No acquisition (no seed M&A flag; none found). Ownership confidence: high.
  • Series F: $150M at $7.2B valuation, announced 2025-06-10, led by Wellington Management (new: Khosla Ventures, Bicycle Capital, Geodesic Capital, Archerman Capital; many existing investors incl. Sequoia, Kleiner Perkins, Lightspeed, ICONIQ, Coatue, DST, General Catalyst, Sapphire, IVP, Capital One Ventures, Citi).
  • Prior: Series E $260M at $4.6B (Sept 2024); ~$2.2B (Feb 2024). TechCrunch puts cumulative funding at ~$610M since early 2024.
  • ARR: surpassed $100M ARR in its last fiscal year, less than three years after launch (vendor/TechCrunch).
  • Founded 2019, HQ Palo Alto, CA. Founders include Arvind Jain (ex-Google, co-founder/CEO).

CTO / hedge-fund lens

  • Day-1 for the RAG/assistant use case. If a fund deploys any enterprise LLM assistant over internal data, entitlement-aware retrieval is the table-stakes control — you cannot let an assistant surface deal docs, LP data, comp, or MNPI to unauthorized staff. Glean is one of the cleaner answers to that requirement.
  • SR 11-7 / model risk: Glean is a retrieval/productivity layer, not a trading or valuation model, so it’s generally outside formal model-risk governance — but its outputs touch sensitive data, so it belongs in data-governance, access-review, and information-barrier (Chinese wall) controls. The permission-mirroring helps enforce info barriers if the source-system ACLs already encode them.
  • Fit: medium for a hedge fund. Strong product, but it is a heavyweight horizontal platform priced for larger enterprises; a 50-person fund may find it oversized versus the platform copilot they already license (M365 Copilot) or a lighter tool. Best fit: larger asset managers with many SaaS silos and a real internal-search pain. Key diligence items: data residency, whether a VPC/isolated tenancy is available, MNPI/information-barrier handling, and how it treats data in model grounding (Glean states it does not train foundation models on customer data — verify contractually).

Competitors / alternatives

  • microsoft-365-copilot — bundled Microsoft-suite copilot; permissions via microsoft-graph.
  • gemini-enterprise — Google Agentspace / Gemini Enterprise.
  • microsoft-graph — permissioned-retrieval substrate (build-vs-buy alternative for Microsoft-centric shops).
  • Other enterprise-search/RAG players (Sinequa, Coveo, Elastic-based builds) compete on retrieval but with varying assistant/agent maturity.

Open questions / to verify

  • Is a customer-managed VPC / self-hosted / on-prem (e.g., “Glean in your cloud”) deployment offered, and at what tier? Primary docs fetched did not confirm.
  • Data residency options (US/EU) and tenancy isolation specifics for a regulated buyer.
  • Exact cumulative funding figure (sources cite ~$610M since early 2024; total-since-inception not pinned).
  • Strength of oversharing prevention when upstream source ACLs are misconfigured (independent research e.g. Knostic raises this) — a real diligence question for any permission-mirroring tool.

Sources

History

  • [2026-06-28] Stub created from seed registry.
  • [2026-06-28] Researched; established founding 2019 / HQ Palo Alto, independent venture-backed, Series F $150M at $7.2B (2025-06-10, lead Wellington), ~$610M cumulative, >$100M ARR; documented entitlement-aware-RAG mechanics (connector ACL sync + query-time enforcement, defers to source-system permissions). No M&A. Set ownership_confidence high, status researched, hedge_fund_fit medium. Open: VPC/self-hosted availability and data residency unconfirmed.

Graph View

Backlinks