Head-to-head · updated 13 September 2026
DataShield vs Great Expectations: the tests passed, so who let the agent read the table?
Great Expectations wrote the vocabulary. Expectations, Expectation Suites, Checkpoints, Data Docs. If you have ever asserted that a column is never null and had the pipeline stop when it was, you have used the idea GX made normal. GX Core is free, Apache-2.0, and it runs where your data already runs. Their own hero calls it "the most popular data quality framework in the world" and more than 11,000 practitioners back that up. We are not going to pick a fight with a good library.
We do not write Expectations. DataShield is not a rules engine, and you should keep GX Core if rules are what you need. We work one layer over: the dataset itself. Every dataset gets a 20-section profile and a trust score. Sensitive fields get classified against 129 field classes. Datasets are tokenized at ingest; agents query tokenized data over MCP; detokenization is a privileged, audited operation. And every governed tool call an agent makes is sealed into a hash chain you can verify without trusting us. Below is the honest split, rows GX wins included.
The short version
Pick DataShield when
- You want to know what is in a dataset, not just whether it passes. Profiling, 129-class PII and PHI classification, and a trust score per domain, without writing a single assertion. How Ontology does it.
- Agents read your data, and somebody will eventually ask you to prove an access log was not edited. Our chain answers with math. Run the verifier.
- You need to pull an agent's authority mid-session and have the very next tool call fail. Not the next token refresh. How Auth does it.
- You liked GX Cloud but you want the whole thing on your own infrastructure, run by a vendor with a published price.
Pick Great Expectations when
- You want declarative tests in your pipeline code. That is the job GX Core was built for, and we do not do it at all.
- Your checks have to run where your data runs: pandas, Spark, SQL, wired into Airflow next to the rest of your DAG.
- Free matters. GX Core is Apache-2.0 with no feature limits, and Fivetran says it stays that way.
- Your team already speaks Expectations. A shared vocabulary across eleven thousand practitioners is worth real money.
Bottom line: GX Core tests the shape of your data. DataShield governs the data itself and keeps proof of what agents did with it. Run both if you like. But if you were a GX Cloud customer, note that the SaaS went away on 1 June 2026 and the roadmap now lives at two other companies.
Feature by feature
Competitor cells describe what Great Expectations, FICO and Fivetran have published as of the date above. If we have mischaracterised something, email support@myorg.ai and we will correct it, credited.
| What matters | DataShield | Great Expectations | Edge |
|---|---|---|---|
| Data quality rules, tests and alerts | None. We do not author Expectations, we have no rules DSL, no anomaly detection, no monitors on your warehouse tables, no freshness SLAs and no incident management. If you want assertions, keep GX Core. | The whole product, and it is good. Declarative Expectations in Python or YAML, grouped into Suites, run as Checkpoints, rendered as Data Docs. Multi-engine: pandas, Spark, SQL. Email alerts on failures shipped in March 2026. | ◇ |
| Dataset profiling and a trust score | Every dataset gets a 20-section analysis profile: completeness, field statistics, patterns, column semantics, relationship graph, quality metrics, compliance governance, transformation lineage, business rules and a source fingerprint. A weighted composite trust score rolls up per entity type, domain and estate, recomputed hourly with per-axis breakdown and trend. | Profiling exists to help draft Expectations, and ExpectAI generates them from profiling plus business context. The output is a test suite, not a standing profile with a score you can trend. | ◆ |
| PII and PHI classification | 129 field classes covering PII, PHI, financial data and secrets, with deterministic and reproducible verdicts. All 18 HIPAA Safe Harbor identifiers are discrete classes carrying a regulation reference. We also fingerprint the source system behind a dataset from 95 built-ins. | Not a feature. Expectations test shape, range, uniqueness and completeness. Nothing in their public material classifies a column as sensitive. | ◆ |
| Tokenization and generalization | Deterministic, join-preserving, vault-reversible tokens applied at ingest, plus quasi-identifier generalization (dates to year, decade or age band; ZIPs to 3 or 4 digits; partial phones, SSNs and emails) with a measured cardinality-reduction score per column. Both are switches you turn on, not defaults. | Not part of the product. | ◆ |
| Audit evidence | SHA-256 hash chain with Ed25519-signed checkpoints that are themselves chained. The verdict is three-valued (clean, attested damage, tampered) and verification names the failure: tampering, insertion, deletion or truncation. Try the verifier. | Data Docs is an HTML report, and GX added Expectation and Data Asset history in April 2026. Useful for debugging. We found no cryptographic tamper evidence in their public docs. | ◆ |
| Agent authorization | Every governed tool call passes a scope ceiling, a consented-tool allowlist, an authority tier and a revocation re-check before dispatch. The call fails closed, and delegation is RFC 8693 token exchange with an enforced scope ceiling. | No access control plane. ExpectAI runs in the customer's own Agent deployment on the customer's own OpenAI credentials, which is a sensible design, but it governs an LLM key, not an agent's authority over data. | ◆ |
| Break-glass | Scoped, time-boxed emergency access for agents. It auto-revokes, is step-up and IP gated, and cannot be quietly deleted from the log. | Not described in their public material. | ◆ |
| GDPR erasure | Crypto-shred of per-subject key material plus ISO 27560 consent receipts. Actor identities in the chain are HMAC-committed, so the evidence still verifies after the subject is gone. | Out of scope for a test framework, and they do not claim it. | ◆ |
| MCP and agents | More than 200 MCP tools across Ontology, Auth, Corpus and Lighthouse, all first-party. Auth issues MCP tool tokens with scope ceilings, and per-call metering is attributed to the agent. | No first-party MCP server. The gx-mcp-server project that exposes GX checks as MCP tools is maintained by one independent developer under MIT, not by GX. | ◆ |
| Deployment and hosting | Self-hosted in your own cloud or data center, or a dedicated single-tenant server we operate. Docker images ship for Auth, Ontology, Corpus and Lighthouse, and a signed deploy manifest is verified by Guardian. Ed25519 audit-signing keys can live in your own KMS or HSM. HMAC tokenization keys sit in your environment today, not in a KMS. | GX Core is a Python library you host yourself, which is about as portable as it gets. GX Cloud was a SaaS control plane with a self-hosted Agent deployment, and it stopped being publicly available on 1 June 2026. | ◈ |
| Maturity and ownership | One company, one roadmap. Auth, Guardian and Lighthouse are live in production (Guardian and Lighthouse since April 2026). SOC 2 not yet certified, and we say so. | Founded 2017, $65M raised, 11,000-plus practitioners. As of May 2026 GX Cloud belongs to FICO and GX Core is stewarded by Fivetran, which is itself mid-merger with dbt Labs. The library is in safe hands. The product line is not one product line any more. | ◇ |
| Pricing | Published model and a scoped instant quote, no sales wall. | GX Core is free under Apache-2.0 with no feature limits. GX Cloud named Developer, Team and Enterprise tiers and never published a dollar figure for the paid ones. Free is hard to beat, so we will not pretend otherwise. | ◇ |
◆ DataShield leads◇ Great Expectations leads◈ comparable
Great Expectations claims are drawn from greatexpectations.io, the GX blog and Fivetran's press release, last checked 13 September 2026. We link them below rather than work from memory.
Three things you get here that you won't get from a test framework
The data typed, not just tested
A passing test tells you a column is well formed. It does not tell you the column is a patient MRN. We label columns against 129 field classes, and the verdict is deterministic, so two runs agree. All 18 HIPAA Safe Harbor identifiers are in there. See the catalog.
Proof that survives an audit
An HTML report proves very little. Anyone with write access can edit it. Ours is a hash chain with signed checkpoints, and the verifier tells you what broke, not just that something did. That is the property EU AI Act Article 12 and HIPAA §164.312(b) reviewers ask about. Try it in your browser, no signup.
Authority that can change mid-flight
An analyst leaves on a Friday. Their agent is 20 minutes into a 40-minute job. With DataShield the next governed tool call is re-checked against current authority and fails closed. A test suite has no opinion about this at all. How Auth does it.
Where Great Expectations is genuinely stronger
Expectations are a good idea, executed well, for nine years. Writing expect_column_values_to_not_be_null and having it fail your Airflow run is worth more than most governance decks. GX Core is free, Apache-2.0, has no feature limits, and runs on pandas, Spark and SQL so your checks live next to your code instead of in somebody's console. Data Docs turns a test run into something a stakeholder can read. Eleven thousand practitioners share the vocabulary, which means your next hire already knows it. We have no equivalent, we are not building one, and if assertions are the gap in your stack you should install GX Core this afternoon.
Here is the push-back, and it is about the last two years rather than the library. GX's own 2026 posts moved from "data quality" to "trust" to "Trustworthy AI Agents: The Data Quality Pillar of the AI Context Layer" in about four months. That is the right instinct. A test suite is genuinely part of whether an agent should be trusted with a table. But it is the smallest part, and the framework has no way to say which columns hold PHI, no way to stop a specific agent reading them, and no artefact an examiner can verify independently. Then, in May 2026, the commercial product was sold to FICO and shut down for public use on 1 June, while the open-source project moved to Fivetran. If you were buying GX Cloud to solve agent trust, the thing you were buying is now two things owned by two other companies, and neither of them is a data governance vendor.
Questions worth asking both of us
These are the questions we would want answered if we were the ones buying. Ask them on every call, ours included.
Can you cryptographically prove an audit log entry wasn't deleted?
DataShield: yes. Each record commits to the one before it, checkpoints are Ed25519-signed and chained, and the verdict separates deletion from truncation and from tampering. Run it against a sample chain at /verify. Great Expectations: Data Docs is a static HTML report and the newer history view is a UI over run records. We found no tamper-evidence mechanism in their public docs. Ask them to show one.
What happens to a revoked agent mid-session?
DataShield re-checks authority on every governed tool call, so revocation lands on the next call rather than the next token refresh. Great Expectations has no access control plane, so there is nothing to revoke. That is not a criticism of a test framework. It is a reason not to buy one for this job.
How does GDPR erasure interact with the audit trail?
DataShield crypto-shreds the per-subject key material and issues an ISO 27560 consent receipt. Actor identities in the chain are HMAC-committed, so the evidence still verifies once the subject is gone. GX does not claim an erasure mechanism, and a validation run that captured example failing values is worth checking on this point. Ask where those sample values are stored and for how long.
GX Cloud shut down on 1 June 2026. What are my options?
Three, honestly. Keep GX Core, which is free and now stewarded by Fivetran, and run it yourself. Talk to FICO, who bought GX Cloud to fold into their Decision Intelligence platform. Or move the governed-data job to a stack you host. We are the third option, and we are only the right one if your real problem is classification, agent access and evidence rather than assertions. If it is assertions, option one is free and we will say so on the call.
Is DataShield a Great Expectations alternative for data quality testing?
Not for testing, no. We have no rules engine, no anomaly detection and no alerting. We are an alternative for the part of the job GX Cloud was starting to reach for: profiling a dataset, scoring its trust, labelling its sensitive fields, and governing what an agent may do with it. Most teams that buy us for that keep their test framework. Ours is the layer under it.
Great Expectations has an MCP server. Don't you both do agents?
Theirs is gx-mcp-server, an MIT project by one independent developer, not a GX product. It lets an agent run validations, which is useful. Ours is where the governed data itself is queried: the tool token carries a scope ceiling, the call is authorized before dispatch, and the decision is sealed into the chain. Different jobs. Ask GX whether their MCP story is first-party and supported, because today it is not.
Does DataShield have SOC 2?
Not yet, and we will not imply otherwise. Auth ships with a public threat model and a verifier anyone can run. Guardian and Lighthouse have been in production since April 2026. Design-partner terms include source escrow, so a small vendor is not a single point of failure. Details on the security page.
- GX's hero: "A powerful, flexible data quality solution… the most popular data quality framework in the world", with a community of "more than 11,000 data practitioners" and GX Core under Apache-2.0. — greatexpectations.io, 13 Sep 2026
- "Great Expectations (GX) and Fair Isaac Corporation (FICO) have completed a definitive agreement to acquire GX Cloud", and "GX Cloud will no longer be publicly available beginning June 1." — Great Expectations blog, 6 May 2026
- "GX Core will continue as an open source, community-driven project. Fivetran intends to support ongoing project maintenance, ecosystem integrations, and community engagement." — Fivetran press release, 13 May 2026
- ExpectAI runs in a self-hosted Agent deployment on the customer's own OpenAI credentials, and "GX Cloud still cannot connect to or extract your underlying data." — Great Expectations blog, 24 Feb 2026
- GX Cloud listed Developer, Team and Enterprise tiers with no dollar figures published for the paid options. — greatexpectations.io/pricing, 13 Sep 2026
- gx-mcp-server, which "exposes Great Expectations data-quality checks as MCP tools for LLM agents", is an MIT-licensed project maintained by an independent developer, not an official GX repository. — GitHub, 13 Sep 2026
Other head-to-heads
DataShield vs Soda
DataShield vs Soda: Soda owns YAML data quality checks, contracts and anomaly detection. DataShield adds PII.
Same marketDataShield vs Anomalo
DataShield vs Anomalo: Anomalo leads unsupervised data quality monitoring. DataShield adds agent.
Same marketDataShield vs Bigeye
DataShield vs Bigeye: Bigeye wins on data observability, anomaly detection and lineage. DataShield adds.
AllEvery comparison
One honest scorecard per vendor.
See both mechanisms run in your browser: break a live audit chain, revoke an agent mid-session, then decide what your test suite still owes you. Demo Center access is free with a work email.
Get free Demo Center accessYou've seen the proof
Ready for a number? Scope your deployment and we'll price it against your own economics.
Get your quote →