Head-to-head · updated 13 September 2026

DataShield vs Okera: the product is gone, so what governs your data now?

Let's get the awkward part out of the way. Right up front. Okera is not a product you can buy. Databricks bought the whole company. That was in May 2023. Okera.com now redirects to databricks.com. You liked the fine-grained access control. Fair enough. Row filters, column masks, tags that drive rules. All of that lives on now. It's Unity Catalog ABAC now. It's good, no question. But it's only as wide as your Databricks estate. That's the real catch.

DataShield is not a lakehouse catalog. We won't pretend to be one. We are the authorization and evidence layer instead. It sits across whatever your agents touch. We tokenize datasets right at ingest. Agents then query tokenized data over MCP. Turning a token back into real data is a privileged step. It's audited every time. We seal each checked call into a hash chain. You can go verify it yourself, right now. Here is the honest split. It includes the rows Databricks wins.

Get free Demo Center access Get an instant quote

The short version

Pick DataShield when

Your agents reach past the lakehouse. Think Postgres, Snowflake, S3, a vendor API. You get one rule plane. One log covers all of it. Not one per platform.
An examiner will ask you to prove a log wasn't edited. Sooner or later, they will. A table the platform wrote, stores, and can rewrite just isn't proof. Ours is a hash chain with signed checkpoints. Run the verifier.
You need to pull an agent's mid-session access. The very next tool call should fail.
You want the rule engine, the vault, and the evidence running on your own system. You hold the keys. See the architecture.

Pick Unity Catalog when

Your data already lives on Databricks, full stop. ABAC is native there. There's no second vendor to buy, deploy, or renew.
You want row filters and column masks, driven by checked tags. These apply to Delta tables at query time. That happens on the fly. That's Okera's old job, done right. We just don't do it.
You need automatic lineage. It runs from a dashboard back through the tables to the source. Databricks captures that. It just runs the compute. We can't see what we don't sit in front of.
Your oversight budget is really a platform budget. Folding it into DBU spend beats a new purchase order.

Bottom line: Unity Catalog controls the data on Databricks, full stop. DataShield controls what an agent can do, anywhere. It keeps proof of it, too. In Databricks shops, we expect to run right alongside it. Is your estate wider than one platform? Does the evidence need to stand on its own? Start here, then.

Feature by feature

Competitor cells describe what Databricks' public site and docs say. We use the date above. If we got something wrong, email support@myorg.ai. We'll fix it and credit you.

What matters	DataShield	Okera / Unity Catalog	Edge
Product status	Sold and supported as DataShield, today. Auth is live now.	Okera is not a solo product any more. Databricks bought the company back in May 2023. Okera.com now redirects straight to databricks.com. The tech ships today as Unity Catalog ABAC.	◆
Fine-grained access control	Scope ceilings and access tiers sit on each tool token. We also generalize quasi-identifiers, as a rule. Dates become year, decade, or age band. ZIPs become 3 or 4 digits. Phones, SSNs, and emails get partial masking too. Each column gets its own cardinality-reduction score. We don't apply row filters to your Delta tables, though.	ABAC is live now. Checked tags drive row filter and column mask rules. These apply to tables, materialized views, and streaming tables alike. One rule can cover a whole catalog at once. DENY rules are still Beta, for now.	◇
Audit evidence	A SHA-256 hash chain, with Ed25519-signed checkpoints. Those checkpoints are themselves chained too. Verification names the exact failure: tampering, insertion, deletion, or truncation. Try the verifier.	Audit logs, inference tables, and MLflow traces all sit as lakehouse tables. The same platform controls them too. We found no claim at all. Nothing is hash-chained. Nothing is signed.	◆
Agent authorization	Each checked tool call passes a scope ceiling. It also carries an access tier, each time. It gets a re-check before it goes out, too. A pulled grant lands mid-session, not after.	Grants and ABAC rules sit on catalog objects. Plus RBAC on models and tools in Agent Bricks too. They don't describe pulling access mid-session at all.	◆
Break-glass	Scoped, time-boxed emergency access, just for agents. It auto-revokes on its own. It can't be quietly deleted from the log, ever.	Not described at all, in any of the docs we read.	◆
GDPR erasure	Crypto-shred of per-subject key material. Plus ISO 27560 consent receipts too. The audit chain stays fully checkable after erasure.	Delete the rows. That's the whole method. What happens to data copied out of checked tables isn't documented. Neither is the effect on your audit history.	◆
Tokenization and data handling	We tokenize datasets right at ingest. Agents query tokenized data over MCP after that. Turning a token back into real data is a privileged step. It's audited, always. Tokens are deterministic, join-preserving, and vault-reversible too.	Masking and redaction apply at query time, by rule alone. We found no vault-reversible token setup at all. We found no checked path back to real data either.	◆
Reach beyond the platform	Platform-agnostic, by design. Auth issues MCP tool tokens. Checked surfaces confirm them on each call, wherever the data sits. Self-hosted. No compute vendor required.	Lakehouse Federation gives checked access to foreign catalogs, yes. But the docs call it read-only. Table-level controls only. Row and column rules stay a Databricks-table story, still.	◆
MCP and agents	Native MCP endpoints sit on Auth, Ontology, and Lighthouse. MCP tool tokens carry scope ceilings, always. Per-call metering ties right back to the agent.	Managed MCP servers for Genie, SQL, and UC functions ship today, already. External MCP clients can connect in too. These register as catalog securables, though. Not per-call access checks.	◈
Lineage	Typed lineage, across the whole catalog. You can trace it hop by hop. Access gets checked at each hop. It covers asset and column records. It also covers correlation IDs in the audit chain. It comes from the pipelines that own the relationships. Not a separate graph store. It's not column-level lineage across a compute platform, though.	Automatic lineage, across tables, notebooks, dashboards, and models alike. Databricks captures it simply because it runs the query.	◇
Deployment	Self-hosted, in your own cloud or data center, your call. Or on a dedicated single-tenant server we run for you. Your keys, always.	Databricks control plane sits on AWS, Azure, or GCP. A customer VPC holds the data plane. An open-source Unity Catalog server exists too, for what it's worth. It carries a smaller feature set, though.	◆
Maturity and pricing	Auth, Guardian and Lighthouse are all live now. Guardian and Lighthouse have run since April 2026, in fact. SOC 2 isn't certified yet, and we say so plainly. Published pricing and an instant quote.	Databricks is a large private firm. It has a deep bench. There's no set price for oversight at all. It rides on DBU spend instead. A rep quotes it.	◇

◆ DataShield leads◇ Okera / Unity Catalog leads◈ comparable

Databricks claims all come from databricks.com and docs.databricks.com. Last checked on 13 September 2026. We link them below, not just a paraphrase from memory.

Three things you get here that you won't get from a lakehouse catalog

A witness that isn't the defendant

The platform that ran the agent shouldn't be the only witness. Not to what it did. Not ever. Our chain lives in a separate control plane instead. It has its own signing keys. Break it, right in your browser. Watch the verifier name the exact failure. That's the property EU AI Act Article 12 reviewers care about most. Try it, no signup.

Authority that can change mid-flight

An analyst leaves the company on a Friday. Their agent is still grinding through a 40-minute job, unaware. DataShield re-checks the next checked tool call against current access, right then. It fails closed. A grant revoked in a catalog doesn't reach a session that's already running. How Auth does it.

Governance that outlives the platform choice

Okera buyers learned this the hard way. You picked a layer, once. Someone else went and bought that layer. Your policy model went with it, like it or not. Run the rule plane and the evidence plane yourself instead. Do it on your own hardware. A vendor's roadmap stops being your migration plan, that way. How we deploy.

Where Unity Catalog is genuinely stronger

Okera's engineers didn't waste those three years. Not at all. Unity Catalog ABAC is live now. It's a cleaner product than the solo one ever was, honestly. Tag a column just once. The rule follows it across a whole catalog. Row filters, column masks, materialized views, streaming tables. The engine handles all of it at query time. It's the same engine that already holds the data. You get lineage for free. Databricks runs the compute, after all. Does your data already live on Databricks? Then turning this on is just a setup change. We'd tell you to go do it, before you even talk to us.

The push-back is about scope, and about proof, both. Their oversight stops where the platform stops, full stop. Lakehouse Federation does reach foreign catalogs. But the docs are plain that it's read-only. Table-level controls only. That's a long way from row and column rules on your Postgres. And the audit trail itself is just a set of tables. The platform writes them. It can rewrite them too. Databricks made our own point better than we could have. In a blog on 10 September 2026, they wrote something close to our point. Each catalog must show it complies on its own. No admin should have to match up audit trails across systems. We agree. We just think the outside witness should sit above the catalog. It shouldn't be one of them.

Questions worth asking both of us

These are the questions we'd want answered if we were buying. Ask them on each vendor call, including ours.

What happened to Okera, and is there still an Okera alternative to buy?

Databricks acquired Okera back in May 2023. They didn't disclose terms, either. Okera.com still redirects to databricks.com. The fine-grained access control survives on as Unity Catalog ABAC, live now. So the real question is just this: what did you want Okera for, really? If it was row and column rules on Databricks tables, then use ABAC. Maybe it was a data access control layer instead. One that didn't need you to standardize on one compute vendor. That option left the market with the acquisition. That gap is the one we fill.

Can you cryptographically prove an audit log entry wasn't deleted?

DataShield: yes, it can. Each record commits to the one right before it. We sign and chain every checkpoint. Verification tells deletion from truncation from tampering, every time. Run it at /verify. Unity Catalog: audit logs, inference tables, and MLflow traces are all lakehouse tables. They're rich, and they're queryable. Databricks doesn't claim they're hash-chained or signed, though, and we went looking hard. Ask them to show you the integrity proof.

What happens to a revoked agent mid-session?

DataShield re-checks access on each checked tool call, always. Revocation bites on the very next call, no delay. Unity Catalog enforces grants and ABAC rules when a query actually runs. That covers a lot, to be fair. We found no description of access tiers for agents, anywhere. We also found no description of a session getting downgraded in flight. Ask how long a compromised agent keeps working after you pull its access.

How does GDPR erasure interact with the audit trail?

DataShield crypto-shreds the subject's key material. Then it issues an ISO 27560 consent receipt. Actor identities in the chain carry HMAC commitments, not names. The proof still checks out, even after the subject is gone. On the Databricks side, deleting rows is fairly simple. But the docs don't cover data already copied out of checked tables. They don't cover what erasure does to your audit history either. Ask both.

Does DataShield replace Unity Catalog?

No. We'd be lying if we said otherwise, so we won't. Unity Catalog is the catalog for your Databricks tables. It's the policy engine too, full stop. We don't apply row filters to Delta tables. We don't do platform lineage either, not yet. Most Databricks customers who talk to us keep Unity Catalog as is. They just add an independent layer. It's for agent authorization and tamper-evident evidence. There's more on that in our Unity Catalog write-up.

Does DataShield have SOC 2?

Not yet. We won't imply otherwise, ever. Auth is live now. It ships with a public threat model and a verifier anyone can run. Guardian and Lighthouse have both been live since April 2026. Design-partner terms include source escrow. That way a small vendor isn't a single point of failure. Details on the security page.

okera.com returns an HTTP 301 redirect to databricks.com. The solo product is gone. So is its site. - okera.com, 13 Sep 2026
"In May 2023, Databricks bought data security group Okera, extending Databricks data governance capabilities." Terms undisclosed. - Wikipedia, Databricks, 13 Sep 2026
Unity Catalog ABAC is live now. Governed tags drive row filter and mask rules. They cover a whole catalog or schema. DENY rules are Beta. - docs.databricks.com, 13 Sep 2026
Unity Catalog hero, verbatim: "Unified governance for data, apps and AI agents." - databricks.com, 13 Sep 2026
Lakehouse Federation gives "governed, read-only access to external data through Unity Catalog foreign catalogs," with table-level access controls. - docs.databricks.com, 13 Sep 2026
Databricks: "each catalog must be able to demonstrate compliance independently without requiring administrators to reconcile audit trails across multiple systems." - Databricks blog, 10 Sep 2026

Other head-to-heads

Direct

Get free Demo Center access

You've seen the proof

Ready for a number? Scope your deployment and we'll price it against your own economics.

Get your quote →