Compare nine data platforms on column-level lineage, AI request lineage, policy enforcement, and hybrid coverage, plus the rule-change test to run first.
By
Billy Allocca
Table of Contents
Best Data Platforms With Lineage and Governance for AI 2026
Last updated: October 27, 2026
The best data platforms for lineage and governance in 2026 capture column-level lineage across both batch pipelines and AI model requests, enforce policy at the point where data is read, and cover on-premises and cloud sources under one model. Catalogs that only document lineage rely on each source system to enforce policy. This guide compares nine platforms on those criteria.
Governance tends to break at the seams between systems. In NexusOne's 2026 State of the Data Estate report, conducted by Enterprise Management Associates (EMA) in September 2026 (n=202) and commissioned by NexusOne, 81% of data and analytics decision-makers said they must update two or more systems when one business rule changes, against 11% who define a rule once [1]. In the same survey, 54% said they govern and operate data and AI with six or more separate tools (n=202) [1]. So the useful question for a buyer is how many places a policy and its lineage have to live, and which platform keeps that number lowest for your estate. The companion guide on data lineage for AI models explains how tracing a model back to its training and retrieval data works step by step.
What Should Built-In Lineage and Governance Include in 2026?
Built-in lineage and governance should answer three questions for every read by a person, a scheduled job, or an AI agent: where the data came from, who was allowed to read it, and what the read produced. A platform that answers the first question well and the other two poorly is a catalog, which is useful, but it covers only part of the job.
A few terms carry most of the evaluation. Data lineage is the recorded path data takes from its source through each transformation to the reports, models, and applications that consume it. Column-level lineage traces individual fields through that path, so you can see that a revenue column in a dashboard derives from two specific source columns. Transformation lineage records the logic applied at each hop, such as the SQL statement or job that produced a table. AI request lineage links a single model or agent request to the identity that made it, the data it retrieved, the model that answered, the policy decision applied, and the response. A policy enforcement point is the component that actually allows, masks, or blocks a read at query time, as distinct from the tool where the policy is written down. OpenLineage is an open specification for lineage events, governed as a graduate project of the LF AI & Data Foundation, that models datasets, jobs, and runs and lets tools attach extra metadata as facets [2].
Criterion | What to test in a proof of concept | Why it matters for AI |
|---|---|---|
Column-level lineage across engines | Run the same transformation in two engines (for example SQL and Spark) and confirm both produce column edges | Agents and models read columns, and a table-level graph cannot show which sensitive field reached a prompt |
Transformation detail | Open a lineage edge and check that the SQL or job code behind it is visible | Root-cause analysis and model reproducibility need the logic as well as the path |
AI request lineage | Send a test agent request and find the log entry that names the identity, data read, model, and policy decision | Auditors ask which data produced a specific AI decision, and pipeline lineage stops before the request |
Policy enforcement point | Change one masking rule and read the column from every engine you run | A policy written in a catalog protects nothing until an engine enforces it |
Hybrid and on-premises coverage | Connect one on-premises source and one cloud source and compare lineage depth for each | Most regulated estates keep their most sensitive data on infrastructure they own |
OpenLineage support | Check whether the platform emits OpenLineage events, consumes them, or both | Consuming OpenLineage lets lineage from Airflow, Spark, and dbt land in one graph without custom connectors |
Catalog openness | Confirm open APIs and, for lakehouse tables, Iceberg REST catalog support | An open catalog keeps lineage and policy metadata portable if you change platforms |
Systems to update when a rule changes | Change one business rule and count the systems someone edits by hand | This count is the clearest single measure of governance drift |
The last row deserves the most weight, because the per-system burden falls on exactly the controls that matter for AI. Among respondents who must implement at least some policies separately in each system (n=146), 67% reimplement access permissions and 57% reimplement classification and sensitive-data tagging [1]. Across all respondents, 72% said at least some policies must be implemented separately per system (n=202) [1]. When a classification tag is set in one tool and the access rule that depends on it is enforced in four others, lineage can show where data went while the enforcement silently diverges.
How Do Nine Platforms Compare on Lineage and Governance?
The nine platforms fall into three groups. Platform-native catalogs (Databricks Unity Catalog, Snowflake Horizon, Google Knowledge Catalog) capture lineage automatically and enforce policy for their own compute. Cross-platform catalogs and governance suites (Microsoft Purview, Atlan, Collibra, Informatica, BigID) document many systems and push policy into some of them. A control plane (NexusOne's approach) enforces one identity and policy model across several engines and logs AI requests at the same layer. The tables reflect vendor documentation as of October 2026, and a cell reads "check current documentation" where the vendor's documentation did not confirm a capability.
Lineage Coverage
Platform | Table, column, and transformation lineage | AI request lineage | Hybrid and on-prem coverage | OpenLineage |
|---|---|---|---|---|
Microsoft Purview | Table lineage from Data Factory copy, data flow, and SSIS activities; column lineage where the source supports it; Unity Catalog column lineage from notebook runs on clusters [3][4] | Microsoft 365 Copilot prompts and responses captured in the unified audit log [5] | Self-hosted integration runtime reports lineage for copies from Oracle, SQL Server, Teradata, and SAP sources [3] | Check current documentation; custom lineage in the classic Data Catalog uses an Atlas-based REST API [6] |
Databricks Unity Catalog | Automatic table and column lineage for Databricks queries, notebooks, jobs, and dashboards; Catalog Explorer retains lineage captured after September 1, 2024 indefinitely, system tables keep one year [7] | Model versions and serving endpoints appear in the lineage graph; inference tables log requests and responses to Delta tables [7][8] | Databricks workspaces; outside systems added through external lineage APIs [9] | Check current documentation; external lineage docs do not mention it [9] |
Snowflake Horizon | Column lineage between table-like objects, plus datasets, feature views, and models; Enterprise Edition or higher; one-year retention [10] | Cortex Agent traces with spans for planning, tool calls, and SQL; agent lineage is object-level [10][11] | Snowflake accounts; external sources and destinations through external lineage [12] | Consumes OpenLineage COMPLETE events through a REST endpoint, generally available September 3, 2026 [12][13] |
Google Knowledge Catalog (formerly Dataplex Universal Catalog) | Automatic lineage for BigQuery and Vertex AI pipelines; column lineage only for BigQuery and Managed Service for Apache Spark jobs [14][15] | Vertex AI dataset and model lineage; request-level logging not covered in lineage docs [14] | Google Cloud services; external pipelines through OpenLineage [16] | Consumes OpenLineage major version 1 events; skips column links above 1,500 per message [16] |
Atlan | Query-history miners build table lineage; column lineage for relational and SQL sources [17] | Registers AI models, versions, and AI applications; runtime request capture not documented [18] | SaaS; offline miners for on-prem and air-gapped sources, intra-system only [17] | Consumes events from Spark, Airflow, Flink, and other producers [17] |
Collibra | Technical lineage covering columns, temporary tables, and transformation code; the CLI lineage harvester reached end of life July 31, 2026, replaced by Edge [19] | Registry for AI agents, models, and use cases linked to training and inference data; runtime prompt capture not documented [20] | Cloud-only lineage product; check current documentation for on-prem sources [19] | Check current documentation |
BigID | Column-level lineage tied to discovery, classification, and ownership [21] | Inventories AI agents and maps what they can access [22] | Discovery across cloud, SaaS, and databases; check current documentation for on-prem lineage depth [21] | Check current documentation |
Informatica (IDMC, CDGC) | Column-to-column data element lineage that shows transformations and calculations [23] | AI system assets with model lineage to training, fine-tuning, and evaluation data [24] | Secure Agent or serverless runtime environments extract metadata, including Oracle sources [25] | Check current documentation |
NexusOne | DataHub column-level lineage across Trino, Spark, and Apache Kyuubi on Iceberg tables [26] | AI Data Control Plane logs every allowed and blocked request with identity, route, and reasoning | Runs identically on-prem, in any cloud, hybrid, or air-gapped | DataHub accepts OpenLineage events through a REST endpoint [27] |
Governance and Enforcement
Platform | Policy enforcement point | Catalog openness | Systems-to-update test to run |
|---|---|---|---|
Microsoft Purview | Data Map and Unified Catalog hold metadata only, and their roles grant no access to underlying data; Microsoft 365 enforcement runs through sensitivity labels and DLP [28][5] | Atlas-based REST APIs in the classic Data Catalog [6]; check current documentation for Unified Catalog | Change a classification and count the source systems where someone still edits the access rule by hand |
Databricks Unity Catalog | ABAC row filters and column masks on Databricks compute, generally available May 13, 2026 [29] | Open-source Unity Catalog under Apache 2.0 at LF AI & Data, compatible with the Iceberg REST catalog API [30] | Change an ABAC policy and read the table from a non-Databricks engine |
Snowflake Horizon | Masking, row access, and tag-based policies inside Snowflake [31] | Horizon Iceberg REST Catalog API for Spark and Trino; a basic Spark configuration does not enforce data protection policies [32] | Change a masking policy and read the column through an external Spark session |
Google Knowledge Catalog | Check current documentation | Lineage API with client libraries [16] | Change a policy on a BigQuery table and read it from an engine outside Google Cloud |
Atlan | Data policies govern query and preview inside Atlan; tag reverse sync updates existing Snowflake tags [33][34] | Open APIs and SDK [17] | Change a tag and confirm the dependent masking policy changes in every warehouse |
Collibra | Protect pushes native policies into Snowflake, Databricks, BigQuery, and AWS Lake Formation [35][36] | Check current documentation | Change a Protect rule and check each source outside those four |
BigID | Policy-driven controls and remediation workflows; enforcement mechanics not documented [22] | Check current documentation | Change a sensitivity label and trace how the change reaches the engine serving the read |
Informatica (IDMC, CDGC) | Policy pushdown to Snowflake, Databricks, Redshift, BigQuery, S3, Fabric, Power BI, and Tableau; de-identification only inside Informatica services [37] | Check current documentation | Change a data filter policy and confirm pushdown on each listed source |
NexusOne | Apache Ranger policies enforced across Trino, Spark, Kyuubi, and S3 per object; DataHub tags generate the policies | Apache Gravitino catalog exposing the Iceberg REST protocol and federating external catalogs | Change one Ranger policy or DataHub tag and confirm every engine and agent sees the change |
Two patterns stand out. Lineage depth is strongest where the vendor also owns the compute, because the engine can record exactly what it read and wrote. Enforcement reach is strongest where policy compiles into the engines you actually run, which is why Collibra Protect and Informatica Data Access Management list the specific sources they push into. Informatica is now part of Salesforce, which completed the acquisition on November 18, 2025 [38], so check roadmap commitments for non-Salesforce sources during the evaluation.
OpenLineage support decides how much of your pipeline lineage you can collect without custom connectors. If your orchestration runs in Airflow, Spark, or dbt outside a vendor's compute, a platform that consumes OpenLineage events (Snowflake, Google Knowledge Catalog, Atlan, and DataHub in the tables above) can place those jobs in the same graph as its native lineage. Read the limits closely: Snowflake accepts only COMPLETE events and captures column mappings on a best-effort basis [12], and Google drops column-level links from any message with more than 1,500 of them [16]. Retention matters for the same reason, since a model trained eighteen months ago cannot be traced through a lineage store that keeps one year.
Which Platforms Capture Lineage for AI Model Requests as Well as Pipelines?
Pipeline lineage is mature across all nine platforms, but lineage for individual AI requests is newer, and most platforms record it only for AI that runs on their own services. The gap shows up in the survey: auditability of AI decisions was the second most cited weakness, with 37% naming it among the things their environment delivers least well, just behind access across systems without moving data at 38% (n=202, up to three choices) [1]. Among respondents who can tell fixed rules from model inference for at least some decisions (n=194), 43% could show exactly which rules and data produced a decision in all cases [1].
Platform | What is recorded for an AI request | Scope |
|---|---|---|
Databricks Unity Catalog | Request and response payloads in inference tables, token counts in usage system tables, PII guardrails [8]; Unity AI Gateway, generally available August 5, 2026, governs models, agents, and MCP servers [39] | Databricks-hosted and external models served through Databricks |
Snowflake Horizon | Traces per agent turn with tool inputs and outputs, stored in an immutable observability events table [11] | Cortex Agents |
Microsoft Purview | Prompts, responses, and referenced Microsoft 365 files in the unified audit log; Endpoint DLP can warn or block pastes into third-party AI sites from onboarded Windows devices [5] | Microsoft 365 Copilot and supported AI apps |
Google Knowledge Catalog | Lineage for Vertex AI pipeline datasets and models [14] | Vertex AI |
Atlan, Collibra, Informatica | Registry links between AI models or use cases and their data, with approval workflows [18][20][24] | Documentation of AI assets; runtime capture not documented |
BigID | Agent inventory, permissions, and sensitive-data exposure [22] | Access posture rather than per-request logs |
NexusOne | Each request's identity, PII and jailbreak screening result, route, model, cost, and allow or block reasoning | Every model and agent request that passes through the AI Data Control Plane |
MCP, the Model Context Protocol, is an open protocol that lets AI agents discover and call tools and data sources through one interface, and it is how many agents now reach enterprise data. A request that starts in a Microsoft Copilot session, retrieves from Snowflake, and calls a model hosted on Databricks leaves three partial records in three systems. Regulation is moving toward expecting one record: Article 12 of the EU AI Act requires that high-risk AI systems "technically allow for the automatic recording of events (logs) over the lifetime of the system" [40]. If your agents cross platforms, test whether the platform you choose can log requests that run on someone else's compute, because that is where most shortlists diverge.
A request log that can stand in for AI request lineage should hold these fields for every call, whichever platform produces it:
The identity of the person, service, or agent that made the request, and the identity it acted on behalf of.
The datasets, tables, or documents retrieved, ideally down to the column.
The policy decision: allowed, masked, or blocked, with the rule that applied.
The model and version that answered, and where it ran.
The prompt, the response, and a timestamp, retained for as long as your audit window requires.
The token count and cost, attributed to the requester.
If a vendor's log covers the first four only for its own models, plan for a second log wherever requests leave that vendor.
When Is a Catalog Alone Enough?
A catalog alone is enough when one platform enforces nearly every policy that matters, because the catalog then documents rules a single engine already controls. That describes an estate standardized on one warehouse or lakehouse, with AI workloads running on the same vendor's services. It also describes programs whose main need is stewardship: business glossaries, ownership, data quality scores, and evidence for an audit report. Collibra, Atlan, Informatica, and Purview are strong at that work, and the native catalogs in Databricks, Snowflake, and Google Cloud add enforcement for their own compute at no extra integration cost.
The strongest argument for a catalog-only approach is cost and focus. Adding an enforcement layer means another system to run, and if nearly all of your sensitive reads happen in one platform, its native policies already cover them. That argument holds until reads cross engines. In the survey, only 27% said they can enforce one policy definition fully everywhere, and another 51% said mostly, with exceptions (n=202) [1]. Most organizations also built their governance stack from parts: 78% assembled it by combining platforms and tools, and 5% built it in-house (n=202) [1]. Each part adds a place a rule can drift.
Ask two questions before settling on a catalog. Does any AI agent or engine read your governed data outside the platform that enforces the policy? Does any regulated dataset live on infrastructure the catalog can document but cannot enforce on? If both answers are no, a catalog is the right purchase. If either is yes, you need an enforcement point that sits where those reads happen.
How Do You Choose by Estate Type?
Match the platform to where your data lives and where your AI requests run, because lineage depth and enforcement reach both follow the compute.
Estate type | Strongest fit | What to verify | When to look elsewhere |
|---|---|---|---|
Single cloud, one dominant platform | That platform's native catalog: Unity Catalog, Snowflake Horizon, Google Knowledge Catalog, or Purview for a Microsoft estate | Retention windows (Snowflake keeps one year of lineage; Unity Catalog system tables keep one year) [10][7] | A second engine or an outside agent starts reading the same tables |
Multi-cloud, several warehouses | Atlan, Collibra, or Informatica for cross-platform lineage and stewardship | Which sources receive policy pushdown, and whether OpenLineage events from your orchestrator are consumed | Enforcement must cover sources outside the pushdown list |
On-premises plus cloud | Purview with a self-hosted integration runtime or Informatica runtime environments for lineage; a control plane for enforcement on infrastructure you own | Lineage depth for on-premises sources compared with cloud sources | Your on-premises systems run engines no catalog can push policy into |
Regulated (financial services, healthcare, public sector) | Collibra or Informatica for AI registries and assessments, BigID for sensitive-data discovery, plus a request-level log | Whether the platform logs individual AI requests with the data and policy behind each one | Auditors need one trail across clouds and on-premises systems |
Collibra ships an EU AI Act assessment template for AI use cases [20], which helps the documentation side of a regulated program. The runtime side, a log of what each request read and why it was allowed, has to come from whatever sits in the request path. The broader argument for one governance model across both data and AI is in one governance model for data and AI.
How NexusOne Approaches Lineage and Governance for AI
NexusOne is a data and AI control plane that runs across the platforms you already have, and it produces lineage as a byproduct of one identity and policy model shared by people, jobs, and agents. Identity enters once through Keycloak, federated from Active Directory, Okta, LDAP, or SAML. Policy is defined once in Apache Ranger and enforced across Trino, Apache Spark, Apache Kyuubi, and S3 object storage per object, and tagging a dataset in DataHub generates the matching Ranger policies at bucket, schema, and table level. Because every read passes through the same identity and enforcement layer, DataHub's column-level lineage and the audit trail describe the same events. Apache Gravitino exposes the Iceberg REST protocol and federates external catalogs, so lineage and policy reach tables that live in other systems.
For AI requests, the NexusOne AI Data Control Plane screens each request for jailbreaks and PII at the edge, then logs allowed and blocked requests with the reasoning for each decision. It routes deterministic questions to query engines at zero tokens, then to small models, then to frontier models only when needed, and attributes cost to each person or agent with per-role and per-agent budgets and circuit breakers. The request log, the policy decision, and the data lineage share one identity, which is the record an auditor asks for. The design is covered in more depth in the enterprise AI governance control plane.
The layer runs identically on-prem, in the cloud, hybrid, or air-gapped on a customer-controlled Kubernetes cluster, built on 85+ integrated open-source components. NexusOne's published deployment model is 5-5-5-5: about 5 minutes to provision, 5 hours to deploy, 5 days to the first governed workload, and 5 weeks to enterprise production, with Embedded Builders working alongside your team. A top-three US bank that had run three legacy Hadoop vendors connected 30 applications in under 4 weeks, eliminating more than $100 million in license and hardware costs.
NexusOne is the wrong fit if your estate runs on one cloud platform with one engine and your main need is stewardship and documentation; that platform's native catalog, or Atlan or Collibra, will serve you better. To map your own estate against these criteria, book an expert consultation.
Key Takeaways
Built-in lineage and governance should cover where data came from, who could read it, and what the read produced, for people, jobs, and AI agents alike.
81% of data and analytics decision-makers update two or more systems when one business rule changes (n=202), so count that number for every platform you evaluate.
Lineage is deepest where the vendor owns the compute; enforcement reaches only the engines a platform can push policy into.
AI request lineage is new and mostly limited to AI running on the vendor's own services; test cross-platform requests explicitly.
A catalog alone is enough when one platform enforces nearly every sensitive read; once agents or engines read outside it, you need an enforcement point in the request path.
Frequently Asked Questions
What's the Best Cloud Data Platform With Built-In Data Lineage and Governance Tools?
For a single-cloud estate, the strongest choice is usually your main platform's native catalog. Databricks Unity Catalog captures table and column lineage automatically and enforces ABAC row filters and column masks; Snowflake Horizon adds column lineage, tag-based policies, and OpenLineage ingestion; Google Knowledge Catalog covers BigQuery and Vertex AI. Each governs its own compute best, so test reads from outside engines.
What's the Best Data Platform Company for Metadata-Driven Governance and End-to-End Lineage?
For metadata-driven governance across many systems, Atlan, Collibra, and Informatica lead: all three build cross-platform lineage graphs, and Collibra and Informatica push policies into named warehouses. Lineage that spans from source pipelines to AI requests needs an enforcement and logging point in the request path as well, such as a data and AI control plane like NexusOne.
What's the Best Data Platform Provider for Compliance-Focused Data Governance and Lineage in Analytics Pipelines?
Compliance programs usually pair a stewardship catalog with runtime evidence. Collibra and Informatica provide AI registries, approval workflows, and policy pushdown; BigID adds sensitive-data discovery tied to lineage; Purview covers Microsoft 365 audit. Confirm the platform can show, for a single decision, which data was read and which policy allowed it, because EU AI Act Article 12 requires automatic event logs for high-risk AI systems.
Which Data Platform Provides Full Lineage for Both Traditional Data Pipelines and AI Model Requests?
Databricks and Snowflake record both for workloads on their own services, through inference tables and Cortex Agent traces. Lineage that covers pipelines and AI requests across several platforms requires a layer every request passes through. NexusOne logs each model and agent request with identity, data, route, and policy reasoning, and records pipeline lineage in DataHub under the same identity.
What Is the Difference Between Data Lineage and AI Request Lineage?
Data lineage records how datasets are produced: sources, transformations, and the columns that feed each output. AI request lineage records a single model or agent interaction: who asked, what data was retrieved, which model answered, which policy applied, and what came back. You need both to explain an AI decision, and they are most useful when they share one identity model.
References
More Tools, Less Control: The 2026 State of the Data Estate Report, NexusOne, published September 30, 2026 (survey conducted by Enterprise Management Associates (EMA), September 2026, n=202 data and analytics decision-makers at organizations with 500+ employees in the US, Canada, the UK, and France). https://nexus0ne.substack.com/p/more-tools-less-control-the-2026 (retrieved October 20, 2026)
About OpenLineage, OpenLineage (LF AI & Data Foundation). https://openlineage.io/docs/ (retrieved October 20, 2026)
How to connect Azure Data Factory and Microsoft Purview, Microsoft Learn. https://learn.microsoft.com/en-us/purview/data-map-lineage-azure-data-factory (retrieved October 20, 2026)
Connect to and manage Azure Databricks Unity Catalog in Microsoft Purview, Microsoft Learn. https://learn.microsoft.com/en-us/purview/register-scan-azure-databricks-unity-catalog (retrieved October 20, 2026)
Microsoft Purview data security and compliance protections for Microsoft 365 Copilot and other generative AI apps, Microsoft Learn. https://learn.microsoft.com/en-us/purview/ai-m365-copilot (retrieved October 20, 2026)
Microsoft Purview: Build custom lineage with the REST APIs, Microsoft Learn. https://learn.microsoft.com/en-us/purview/legacy/how-to-purview-custom-lineage-api-user-guide (retrieved October 20, 2026)
Lineage in Unity Catalog, Databricks Documentation. https://docs.databricks.com/aws/en/data-governance/unity-catalog/data-lineage.html (retrieved October 20, 2026)
Configure AI Gateway on model serving endpoints, Databricks Documentation. https://docs.databricks.com/aws/en/ai-gateway/configure-ai-gateway-endpoints (retrieved October 20, 2026)
External lineage, Databricks Documentation. https://docs.databricks.com/aws/en/data-governance/unity-catalog/external-lineage (retrieved October 20, 2026)
Data Lineage, Snowflake Documentation. https://docs.snowflake.com/en/user-guide/ui-snowsight-lineage (retrieved October 20, 2026)
Monitor Cortex Agent requests, Snowflake Documentation. https://docs.snowflake.com/en/user-guide/snowflake-cortex/cortex-agents-monitor (retrieved October 20, 2026)
External lineage, Snowflake Documentation. https://docs.snowflake.com/en/user-guide/external-lineage (retrieved October 20, 2026)
Sep 03, 2026: External lineage (General availability), Snowflake Release Notes, September 3, 2026. https://docs.snowflake.com/release-notes/2026/other/2026-09-03-external-lineage-ga (retrieved October 20, 2026)
Use data lineage with Google Cloud systems, Google Cloud Documentation. https://docs.cloud.google.com/dataplex/docs/use-lineage (retrieved October 20, 2026)
Knowledge Catalog FAQ (renamed from Dataplex Universal Catalog on April 10, 2026), Google Cloud Documentation. https://docs.cloud.google.com/knowledge-catalog/docs/faq (retrieved October 20, 2026)
Integrate with OpenLineage, Google Cloud Documentation. https://docs.cloud.google.com/dataplex/docs/open-lineage (retrieved October 20, 2026)
Capture, read, and troubleshoot lineage, Atlan Documentation. https://docs.atlan.com/product/capabilities/lineage/best-practices/capture-and-troubleshoot-lineage (retrieved October 20, 2026)
AI governance, Atlan Documentation. https://docs.atlan.com/product/capabilities/governance/ai-governance (retrieved October 20, 2026)
Collibra Data Lineage, Collibra Product Resources (2026.02). https://productresources.collibra.com/docs/collibra/2026.02/Content/CollibraDataLineage/co_collibra-data-lineage.htm (retrieved October 20, 2026)
About Collibra AI Governance, Collibra Product Resources (2026.08). https://productresources.collibra.com/docs/collibra/2026.08/Content/AIGovernance/co_about-ai-governance.htm (retrieved October 20, 2026)
Data Lineage, BigID. https://bigid.com/solutions/data-lineage/ (retrieved October 20, 2026)
AI Agent Governance, BigID. https://bigid.com/solutions/ai-agent-governance/ (retrieved October 20, 2026)
View data lineage, Informatica Metadata Command Center Documentation. https://onlinehelp.informatica.com/IICS/prod/MCC/en/metadata-command-center-informatica-intelligent-cloud-services-sources/View_data_lineage.html (retrieved October 20, 2026)
Build Trustworthy AI on Databricks with Informatica AI Governance for Databricks Models, Informatica Blog, February 26, 2026. https://www.informatica.com/blogs/build-trustworthy-ai-on-databricks-with-informatica-ai-governance-for-databricks-models.html (retrieved October 20, 2026)
Introduction to Oracle catalog sources, Informatica Metadata Command Center Documentation. https://onlinehelp.informatica.com/IICS/prod/MCC/en/metadata-command-center-catalog-source-configuration/Introduction_to_Oracle_catalog_sources.html (retrieved October 20, 2026)
About DataHub Lineage, DataHub Documentation. https://docs.datahub.com/docs/features/feature-guides/lineage (retrieved October 20, 2026)
OpenLineage, DataHub project documentation (GitHub). https://github.com/datahub-project/datahub/blob/master/docs/lineage/openlineage.md (retrieved October 20, 2026)
Data governance overview, Microsoft Purview, Microsoft Learn. https://learn.microsoft.com/en-us/purview/data-governance-overview (retrieved October 20, 2026)
ABAC row filtering and column masking policies, governed tags, and data classification are now generally available, Databricks Blog, May 13, 2026. https://www.databricks.com/blog/abac-row-filtering-and-column-masking-policies-governed-tags-and-data-classification-are-now (retrieved October 20, 2026)
Open Sourcing Unity Catalog, Databricks Community, June 18, 2024. https://community.databricks.com/t5/announcements/open-sourcing-unity-catalog/td-p/74815 (retrieved October 20, 2026)
Attribute-based access control (ABAC) using tag-based policies, Snowflake Documentation. https://docs.snowflake.com/en/user-guide/tag-based-policies (retrieved October 20, 2026)
Access Apache Iceberg tables with an external engine through Snowflake Horizon Catalog, Snowflake Documentation. https://docs.snowflake.com/en/user-guide/tables-iceberg-access-using-external-query-engine-snowflake-horizon (retrieved October 20, 2026)
Data policies, Atlan Documentation. https://docs.atlan.com/product/capabilities/governance/access-control/data-policy (retrieved October 20, 2026)
Manage Snowflake tags, Atlan Documentation. https://docs.atlan.com/apps/connectors/data-warehouses/snowflake/how-tos/manage-snowflake-tags (retrieved October 20, 2026)
Set up Protect, Collibra Product Resources (2026.08). https://productresources.collibra.com/docs/collibra/2026.08/Content/Protect/ta_install-protect.htm (retrieved October 20, 2026)
Snowflake policies, Collibra Product Resources (2026.02). https://productresources.collibra.com/docs/collibra/2026.02/Content/Protect/co_snowflake-policies.htm (retrieved October 20, 2026)
Data access policy enforcement methods, Informatica Data Governance and Catalog Documentation. https://onlinehelp.informatica.com/IICS/prod/DGC/en/ae-data-accessmanagement/Data_access_policy_enforcement_methods.html (retrieved October 20, 2026)
Salesforce completes acquisition of Informatica, Seeking Alpha, November 18, 2025. https://seekingalpha.com/news/4523363-salesforce-completes-acquisition-of-informatica (retrieved October 20, 2026)
Announcement: Unity AI Gateway is generally available, Databricks Community, August 5, 2026. https://community.databricks.com/t5/announcements/announcement-unity-ai-gateway-is-generally-available/td-p/164943 (retrieved October 20, 2026)
Article 12: Record-keeping, Regulation (EU) 2024/1689 (AI Act), AI Act Service Desk, European Commission. https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-12 (retrieved October 20, 2026)
