About
Governed intelligence your auditors can approve.
ETSquare is an enterprise AI platform built for work that has to defend itself — in front of auditors, regulators, compliance teams, and general counsel. It combines governed retrieval, enforceable policy, and verifiable auditability in one enterprise runtime.
Why this exists
Pilots that work. Production that won’t clear legal.
The pattern is familiar. A team runs a promising AI pilot. Results look good. Then it meets the compliance reviewer, the security team, and the audit partner — and stops there. Nobody can explain where the answer came from. The policy controls are advisory. The logs are incomplete. The model is opaque. The pilot dies at the production gate.
ETSquare was built for that last mile. The belief is simple: if an AI system cannot be audited, it cannot be trusted with financial, legal, or regulatory work. Every design decision in the product follows from that belief.
What we do
Start with filings. Expand into the enterprise.
ETSquare started with SEC filings because public disclosures are the highest-trust data source in finance — certified, time-stamped, and filed under penalty of law. Every answer, XBRL metric, KPI, insider transaction, and 13F holding surfaced by the product links back to the exact filing on SEC.gov. Nothing is invented. A deterministic retrieval pipeline keeps citations verifiable end to end. The corpus covers over four million validated filing sections across 5,600+ issuers, with normalized XBRL metrics and disclosure patterns built for analyst and agent workflows.
From there, the same governed runtime extends into internal enterprise work: approved documents, ERP data, and operational queries. The trust model does not change between external and internal — only the corpus does. The product is designed to live alongside existing enterprise tools through an API and an MCP server, so internal copilots and analyst chat surfaces can use ETSquare as their evidence and control layer.
How it’s different
Four commitments that shape everything.
Evidence-linked by design.
Every important claim traces to a primary source. SEC answers link to the filing URL and chunk. Internal answers link to the compiled query, policy verdict, and execution trace. “Trust me” is not an acceptable output.
Policy enforced, not advised.
Prompt guardrails are suggestions. ETSquare compiles policy into the retrieval path itself, so unsafe actions are blocked before execution — not explained after. Policy changes without redeployments.
Replayable, auditable decisions.
Every request captures the tool chain, policy verdicts, query lineage, and identity scope into tamper-evident audit records. Audit workflows can verify that nothing was altered after the fact.
Retrieval-only. No fine-tuning.
Customer data stays in the governed database. The model endpoint is replaceable — Claude, GPT, or a customer-hosted model. Citation integrity, tenant isolation, and model freshness are preserved without locking into a single provider.
Defensibility
Two patent applications. One sovereign runtime.
ETSquare’s retrieval and governance architecture is the subject of two US patent applications — 19/400,228 (governed data plane: compiled retrieval, row-level policy, tamper-evident audit) and 19/403,821 (federated governance nodes with bounded authority). Both are implemented in the product and form the technical basis for audit-grade behavior.
The runtime is designed to be deployed in a customer-controlled environment, including in-tenant installations on a managed enterprise cloud database. Deployment details and security model are available under NDA.
Facts, not opinion.
SEC filings are certified disclosures filed under penalty of law. Every data point traces to a sworn statement by a corporate officer. We build on that foundation because it is the only foundation that survives legal, audit, and regulatory scrutiny. We do not publish price targets, analyst takes, or forward-looking opinions. We surface what the filings actually say, with citations you can verify yourself.
The same discipline applies to internal intelligence. If a claim cannot be defended by evidence, the platform is designed not to make it.
Talk to us.
ETSquare is in pilot access with selected enterprise teams. If you are evaluating AI for regulated finance, compliance, or internal audit workflows — and your security review has been the blocker — we would like to hear from you.
Patent applications pending. Third-party product names, trademarks, and service marks belong to their respective owners and are referenced for compatibility only, without implying affiliation, endorsement, or sponsorship.