Redstrata Knowledge Base

Collibra

13 min read

Collibra

Type: company (data & AI governance platform vendor) Sector: Data governance / catalogue / AI governance software First seen: 2026-06-21 Last updated: 2026-09-11

Updated 2026-09-11 based on S-2026-09-11-weekly-vendor-synthesis (weekly vendor-synthesis): the 2026.08/2026.09 lifecycle notices (Console log End of Life; Dataplex-ingestion freeze; breaking filter-API change) are classed as a further instance of Vendor Lifecycle Events as Evidence-Continuity Risk — the second from Collibra in six weeks after the 31 Jul CLI harvester EOL — with the specific read that the audit-log retrieval path is itself DORA / operational-resilience evidence and the Console-EoL-to-Log-API interval is undated. Client check named by the synthesis: Log API availability before cut-over, and Edge-equivalence for GCP estates. Also noted: Collibra broke the pure-play quiet streak but with lifecycle notices, not features (the 2026.09 feature notes remain uncaptured); no Collibra contribution to Apache Ossie recorded while Ataccama shipped one. Per-item facts unchanged. Created 2026-06-21 from S-2026-06-16-collibra-databricks-governance and S-2026-06-02-collibra-snowflake-ai-command-center (daily vendor-intelligence scan). Single-vendor-source so far — capability claims are Collibra’s own and not independently verified. Updated 2026-09-08 based on S-2026-09-04-collibra-202609-announcements (catch-up data-governance vendor-intelligence scan for the 29 Aug–7 Sep 2026 outage window): platform lifecycle notices from releases 2026.08 (21 Aug) and 2026.09 (4 Sep) — Google Dataplex ingestion frozen to defect-fixes with documentation removed, End of Life for log access via Collibra Console (migrate before 2027), a breaking rework of the derived-relation-types filter API, and a Collibra Units vector-storage price cut. Integration/evidence-continuity signals rather than features; the 2026.09 feature notes themselves did not render and remain uncaptured. Updated 2026-07-06 based on S-2026-07-06-collibra-release-announcements (July 2026 release wave: DQ&O AI suggested rules GA, 2026.07 integration-behaviour changes, and 31 July EOL for the CLI lineage harvester). Updated 2026-07-13 based on a full-text refetch of S-2026-06-16-collibra-databricks-governance — added AI Trust Score operational signals, access governance/masking in Databricks, Deasy Labs unstructured-data integration, preview-only availability, and a Lakebase integration stated for Oct/Nov 2026. Updated 2026-08-28 based on S-2026-08-10-collibra-banking-readiness-gap — FS-sector positioning, not a product signal: a 10 Aug blog argues banking’s AI deployment is outrunning its governance capability, frames ungoverned models as compounding “AI governance debt”, and prescribes a single “governing semantic model” above platform-specific semantic layers plus governance embedded in workflow. Notable as the first captured instance of Collibra marketing explicitly against the extended EU AI Act high-risk timetable. All regulatory facts in the post are second-hand and unverified — see Open Questions. Updated 2026-07-27 based on S-2026-06-28-collibra-202606-release-notes (primary 2026.06 release notes: MCP server external access at GA via OAuth 2.0, AI-agent operating model extension, AIUC-1 lifecycle wiring, lifecycle management for all asset types, Vertex AI foundation-model lineage; also surfaced a release-stage tension on AI-suggested DQ rules — see Tensions).

Snapshot

Collibra is a data and AI governance platform vendor (catalogue, lineage, policy, data quality) that now positions itself as an “enterprise AI control plane.” It is relevant to the wiki because catalogue/governance platforms are the layer through which EU/UK regulated firms operationalise BCBS 239 lineage, GDPR usage policy and EU AI Act oversight — and Collibra is pushing hard into governing agentic AI [S-2026-06-16-collibra-databricks-governance][S-2026-06-02-collibra-snowflake-ai-command-center].

Positions / Claims they advance

  • Markets an AI Command Center as a “control layer” for agentic AI, claiming real-time visibility and end-to-end traceability “spanning data sources, models, decisions, and outcomes” [S-2026-06-02-collibra-snowflake-ai-command-center].
  • Provides an MCP Server that delivers governed context (certifications, quality scores, usage policies) to AI agents “in real time, before they act” [S-2026-06-02-collibra-snowflake-ai-command-center].
  • Expanded a bi-directional integration with Snowflake (Horizon Catalog, Cortex Analyst, Cortex Agents) and joined the vendor-neutral Open Semantic Interchange (OSI) standard [S-2026-06-02-collibra-snowflake-ai-command-center]. Cross-reference (2026-09-10): OSI has since been renamed Apache Ossie (Incubating) on entering the Apache Incubator; the first watchlist vendor recorded contributing a converter to it is Ataccama (DQ signals into the semantic layer, announced 9 Sep 2026) — no Collibra contribution to Ossie has been captured [S-2026-09-09-ataccama-apache-ossie-converter].
  • Named Databricks’ 2026 Data Governance Partner of the Year; expanded a bi-directional Unity Catalog integration with capabilities spanning Genie and Agent Bricks, capturing runtime lineage from Databricks into the AI Command Center [S-2026-06-16-collibra-databricks-governance].
  • Claims the AI Command Center’s AI Trust Score now ingests live Agent Bricks operational signals — daily LLM-judge pass rates on groundedness/safety, AI UC-1 assessment status, token-consumption trends — with a fleet-wide dashboard flagging unmonitored agents; the Databricks integrations are preview-only (account-team request), and a Collibra Insights → Databricks Lakebase integration is stated for October/November 2026 [vendor claims — S-2026-06-16-collibra-databricks-governance, full text].
  • Extends into unstructured data via Deasy Labs (a Collibra company): qualifies files as “relevant, unique, fresh, and safe” and writes a learned taxonomy into Unity Catalog as volume schema [vendor claim — S-2026-06-16-collibra-databricks-governance, full text].
  • Now brands itself “the enterprise AI control plane” in company boilerplate [S-2026-06-16-collibra-databricks-governance, full text].
  • Argues that “an agent that acts on stale definitions, uncertified data, or incomplete lineage is an agent operating outside enterprise controls” — i.e. governed context is a precondition for safe production agents [S-2026-06-16-collibra-databricks-governance].
  • Ships GenAI-assisted data-quality rule creation: “AI suggested rules” (technical DQ rules generated from governance policies) and a DQ program dashboard reach GA in July 2026 in Data Quality & Observability [search-derived, not directly verified — S-2026-07-06-collibra-release-announcements].
  • Is retiring legacy lineage tooling: the standalone CLI lineage harvester and single-file custom lineage definition reach End of Life on 31 July 2026, with technical lineage via Edge as the successor path [S-2026-07-06-collibra-release-announcements].
  • Continues consolidating integration paths onto Edge and retiring Console-era interfaces (Aug–Sep 2026): with release 2026.09 (4 Sep 2026) Google Dataplex ingestion is “no longer in active development” (defect fixes only, documentation removed) and the Google Knowledge Catalog integration via Edge is the recommended GCP metadata path; with 2026.08 (21 Aug 2026) the Console API dgcLog endpoints and Console-based Log Proxy reach End of Life “later this year”, to be replaced by a forthcoming Collibra Log API and Log Proxy 1.4.0, with migration required “before 2027”; and the derived-relation-types (preview) filter API was reworked such that CMA package migration across the 2026.08 boundary fails in both directions [S-2026-09-04-collibra-202609-announcements]. Separately, Collibra Units vector-storage pricing for Collibra AI was cut effective 3 Sep 2026 with a retroactive credit [S-2026-09-04-collibra-202609-announcements]. No rationale is given for the Dataplex decision. For regulated users these are dated migration obligations affecting lineage-ingestion coverage and audit-log retrieval — both evidence paths — rather than capability gains [inference from the notices — S-2026-09-04-collibra-202609-announcements].
  • Made the MCP server externally consumable at GA in release 2026.06 (all environments 28 June 2026): OAuth 2.0 authorization code grant lets MCP-compliant clients — the release notes name Claude Desktop, Databricks, Microsoft Copilot and Snowflake Cortex — query Collibra for definitions, ownership, lineage and compliance information [S-2026-06-28-collibra-202606-release-notes]. This turns the earlier MCP-server marketing claim [S-2026-06-02-collibra-snowflake-ai-command-center] into a dated, GA platform capability.
  • Extends the AI Command Center / AI Governance operating model for agent governance: AI Agent Version and AI Agent Tool asset types plus an AI Agent Deployment complex relation (private preview), and the out-of-the-box AIUC-1 Compliance Assessment wired into the AI Agent lifecycle — mandatory at “Under Review”, optional at “Accepted” (GA) [S-2026-06-28-collibra-202606-release-notes].
  • Extends lifecycle management to all asset types (public preview) with required activity gates (assessments, sign-offs) before stage transitions, and introduces smart checks (public preview) — real-time metadata-quality validation planned to replace Validation Rules and Articulation Score [S-2026-06-28-collibra-202606-release-notes].
  • FS-sector market positioning (thought-leadership, 10 Aug 2026): argues banking AI deployment is outrunning readiness “beyond a factor of three” and that regulators on both sides of the Atlantic “just conceded they can’t keep pace” — citing US SR 26-2 as excluding generative/agentic AI from codified model-risk standards, and the EU Digital Omnibus as deferring the AI Act high-risk deadline to 2 December 2027. Its conclusion is that deferral relocates rather than removes accountability: “Every ungoverned agent or model put in production this year is AI governance debt” [S-2026-08-10-collibra-banking-readiness-gap]. All regulatory statements are Collibra’s second-hand characterisations and were not independently verified — they must be cross-checked against primary regulator sources before use.
  • Product thesis advanced in the same post: that “one semantic model will rule above the many” — a “governing semantic model that sits above the data platforms and niche tools” — and that governance must be “embedded in the workflow, not bolted on … part of how AI is designed, approved, deployed, monitored, tested, validated and changed” [vendor thesis — S-2026-08-10-collibra-banking-readiness-gap]. This is the marketing frame for Collibra’s own platform position; the accompanying diagnosis (that the skills gap is “a governance gap in disguise” — “You can buy a model. You cannot buy the processes and institutional knowledge”) is the more transferable part.
  • Notes a divergence useful for readiness framing: for a business unit “ready means a use case works”, while for risk it means “a clear line of sight to explainability, auditability and accountability, which is often the same evidence-based approach regulators expect” [S-2026-08-10-collibra-banking-readiness-gap].
  • The Vertex AI integration now establishes lineage from customer fine-tuned/distilled models back to the vendor foundation models they derive from (Model Garden), capturing the publishing vendor and model version [S-2026-06-28-collibra-202606-release-notes] — a candidate mechanism for EU AI Act model-provenance evidence [inference].

Relationships

  • partners-with → Immuta — both extend governance/access controls into the same Snowflake (Horizon/Cortex) agentic stack; complementary catalogue vs. access-control layers [inference, based on overlapping Snowflake Summit 26 launches].
  • competes-with → Informatica — overlapping catalogue/governance/MDM and agentic-governance positioning [inference].
  • competes-with → Atlan — both Leaders in the 2026 Gartner MQ for D&A Governance Platforms; overlapping catalogue/agentic-governance positioning [S-2026-06-04-atlan-gartner-leader-agentic-governance].
  • relates-to → Agentic Data Access Governance — central vendor in this capability category [S-2026-06-16-collibra-databricks-governance].
  • relates-to → BCBS 239 and Data Lineage — lineage/traceability claims bear on BCBS 239 [S-2026-06-16-collibra-databricks-governance].
  • relates-to → EU AI Act — AI lifecycle traceability bears on AI Act oversight/logging [S-2026-06-02-collibra-snowflake-ai-command-center].

Tracked changes

  • 2026-06-02 — At Snowflake Summit 26, announced expanded bi-directional Snowflake integration; positioned AI Command Center as control layer; joined OSI [S-2026-06-02-collibra-snowflake-ai-command-center].
  • 2026-06-16 — Named Databricks 2026 Data Governance Partner of the Year; expanded Unity Catalog / Genie / Agent Bricks integration [S-2026-06-16-collibra-databricks-governance].
  • 2026-06-23 — Named a Leader in the Nucleus Research 2026 Data Governance Technology Value Matrix (alongside Alation, Atlan, Oracle, Salesforce/Informatica) [S-2026-06-23-nucleus-data-governance-value-matrix].
  • 2026-07-06 — July 2026 release wave logged: DQ&O “AI suggested rules” + DQ program dashboard GA July [search-derived]; 2026.07 platform release (scheduled 19 July) flips Import API continueOnError default to true and changes lastSyncDate behaviour, with /mappings API property deprecations; CLI lineage harvester and single-file lineage definition EOL 31 July 2026 [S-2026-07-06-collibra-release-announcements].
  • 2026-07-13 — Full text of the 16 June Databricks release re-ingested: AI Trust Score operational signals from Agent Bricks, access governance/masking in Databricks, Deasy Labs unstructured-data integration, preview-only availability, Lakebase integration roadmapped Oct/Nov 2026 [S-2026-06-16-collibra-databricks-governance].
  • 2026-07-31 — CLI lineage harvester and single-file custom lineage definition End of Life took effect today; announcements page re-verified by direct fetch — no reprieve or date change. Firms still on the CLI harvester now run unsupported lineage tooling; Edge-based technical lineage is the sole supported path (raw recheck: /_raw_sources/S-2026-07-31-collibra-eol-effective-recheck.md) [S-2026-07-06-collibra-release-announcements].
  • 2026-08-10 (logged 2026-08-28) — Published FS-sector thought-leadership (“The readiness gap”, George Trujillo) positioning against the extended EU AI Act high-risk timetable and advancing the “governing semantic model above the platforms” thesis. No product, release or availability claim; no named EU/UK FS customer [S-2026-08-10-collibra-banking-readiness-gap].
  • 2026-08-21 (logged 2026-09-08 — catch-up) — Release 2026.08 announcements: End of Life for log access through Collibra Console (Console API dgcLog endpoints and Console Log Proxy; Collibra Log API and Log Proxy 1.4.0 as successors; migrate before 2027); derived-relation-types (preview) filter API reworked with a hard CMA-migration boundary at 2026.08 [S-2026-09-04-collibra-202609-announcements].
  • 2026-09-04 (logged 2026-09-08 — catch-up) — Release 2026.09 announcements: Google Dataplex ingestion frozen to defect fixes, documentation removed, Edge-based Google Knowledge Catalog integration recommended; Collibra Units vector-storage price decrease effective 3 Sep 2026 with retroactive credit. 2026.09 feature release notes not retrieved (page did not render) [S-2026-09-04-collibra-202609-announcements].
  • 2026-06-28 (logged 2026-07-27) — Release 2026.06 reached all environments: MCP server external access GA via OAuth 2.0 (Claude Desktop, Databricks, Microsoft Copilot, Snowflake Cortex named as clients); AI Agent Version/Tool asset types + AI Agent Deployment relation (private preview); AIUC-1 assessment mandatory at “Under Review” in the AI Agent lifecycle (GA); lifecycle management for all asset types and smart checks (public preview); Vertex AI Model Garden foundation-model lineage; workflow-engine Groovy command-injection vulnerability fixed [S-2026-06-28-collibra-202606-release-notes].

Tensions

On the release stage of AI-suggested DQ rules:

  • Source A [S-2026-07-06-collibra-release-announcements] (medium, search-derived) records “AI suggested rules” and a DQ program dashboard reaching GA in July 2026 in Data Quality & Observability.
  • Source B [S-2026-06-28-collibra-202606-release-notes] (medium, primary release notes) lists “Suggest Rules” (DEV-164339) as Private Preview in the 2026.06 platform release.
  • Where they actually disagree: possibly nowhere — DQ&O cloud may run a separate release train from the platform release notes, or the feature may have progressed preview→GA between late June and July. But on the face of the two sources the availability status conflicts.
  • Status: unresolved — check the DQ&O cloud release notes directly in a future scan.

Open Questions

  • No EU/UK FS reference customer named in either release — adoption evidence in regulated EU/UK FS is unconfirmed. Reinforced 2026-08-28: the FS-targeted 10 Aug blog also names no bank, insurer or asset-manager customer [S-2026-08-10-collibra-banking-readiness-gap].
  • Is the 2 December 2027 EU AI Act high-risk deadline (via the Digital Omnibus, Council approval stated as 29 June 2026) correct as Collibra states it? This is the single most consequential factual claim in the 10 Aug post for EU/UK readiness planning and needs primary verification against the regulatory scan [S-2026-08-10-collibra-banking-readiness-gap].
  • Does SR 26-2 exist as described, and does it genuinely carve generative and agentic AI out of codified US model-risk standards? If so it marks a real divergence from the EU AI Act approach [S-2026-08-10-collibra-banking-readiness-gap].
  • Collibra’s FS messaging is US-first and silent on UK supervisory expectations (FCA/PRA) despite addressing a UK-relevant audience — positioning gap or author vantage point? [S-2026-08-10-collibra-banking-readiness-gap]
  • What would evidence the existence of a “governing semantic model above the platforms” in a supervisory review? Asserted as necessary, never operationally defined [S-2026-08-10-collibra-banking-readiness-gap].
  • Whether the traceability/lineage claims meet specific BCBS 239 and EU AI Act evidentiary thresholds (unverified).
  • Whether GenAI-suggested DQ rules are explainable/auditable enough to serve as BCBS 239 data-quality controls in a regulated firm [S-2026-07-06-collibra-release-announcements].
  • How many FS Collibra estates still depend on the CLI lineage harvester and face the 31 July 2026 migration-to-Edge deadline for lineage continuity [S-2026-07-06-collibra-release-announcements].
  • For FIs that retrieve Collibra application/audit logs via the Console API or Log Proxy as DORA / operational-resilience evidence, is there a gap between Console log End of Life (“later this year”) and availability of the replacement Collibra Log API? Exact dates not stated [S-2026-09-04-collibra-202609-announcements].
  • Is the Edge-based Google Knowledge Catalog integration feature-equivalent to the frozen Dataplex ingestion for GCP-centric estates using Collibra for BCBS 239 lineage or GDPR data inventory? Not stated; no rationale given for the change [S-2026-09-04-collibra-202609-announcements].
  • What did platform release 2026.09 (4 Sep 2026) actually ship? The feature release notes were not retrievable this run [S-2026-09-04-collibra-202609-announcements].

Sources

Graph View

Backlinks

Recently updated

See 113 more →