EVE AI Core
The Infrastructure of No.
EVE AI Core
The Infrastructure of No.
A capable team can stand up a rules engine, an ALLOW / BLOCK / MODIFY response, Ed25519 signing and an append-only log in a sprint or two. We will say that plainly. The moat isn't any one of those features — it's keeping a governance substrate verifiable, replayable, upgrade-safe, fail-closed and examiner-ready for years, across every key rotation, policy change and infrastructure migration. This page is an honest accounting of that scope.
Standard 60-day pilot: $37,500. Founding design-partner rate: $27,500 for the remaining qualifying cohort slots — credited 100% toward an annual license if you continue.
Let's be direct: a capable engineering team can reproduce the visible surface of a governance control. These parts are real, well-understood, and buildable. If they were the whole product, you would be right to build it.
Evaluate a request against fixed rules and return a verdict. A YAML or JSON policy format on top is a weekend of work.
Three dispositions and a reason string. The API contract is the easy part.
Sign a JSON blob with a private key. Every crypto library ships this.
Write decisions to a hash-chained file. Straightforward on day one.
Front your app with a call to the engine. Standard integration pattern.
Threshold and z-score alerting. A few hundred lines.
That's the demo. The demo is not the product. The next three sections are where a multi-year engineering operation actually lives — and where an internal rebuild gets slower, riskier and more expensive than a license.
Three problems separate a governance demo from a governance substrate. None of them is a feature you finish; each is an operation you run for years.
Signing one decision is trivial. Keeping evidence verifiable across years is not. It means a stable certificate schema across versions, deterministic canonicalization (RFC 8785 / JCS), binding request + response + rule‑result + policy version into one record, key generation / rotation / revocation with historical continuity, cross-language verifier compatibility, tamper detection, replay packages, long-term retention, and Merkle aggregation with inclusion proofs. A certificate signed last year must still verify after you rotate keys and change serializers.
You aren't buying 27 folders of rules — you're buying an operation: regulatory monitoring, version-controlled updates, citation provenance, change-impact analysis, regression testing, customer-specific overrides, effective dates, rollback, and the ability to replay a past decision under the exact policy version in force at the time — with evidence of who approved each change. That's a standing policy-engineering function, not a one-time port.
A prototype passes the happy path. A control has to stay correct during governance-service outages, database outages, duplicate requests, concurrent authorization consumption, network retries, key-service failure, policy-load failure, partial deploys, multi-region failover, version mismatch, clock drift, a direct bypass attempt, and a payload tampered after approval. EVE binds the exact approved payload, resists replay, and consumes authorization at‑most‑once.
One boundary separates the two. Everything on the left is yours to own and differentiate; everything on the right is a maintained substrate you'd otherwise build — and defend — for years. A license buys the right-hand column.
Your competitive surface. EVE never touches it.
The substrate an internal rebuild would own indefinitely.
Right-column items map to shipping components; anything in preview is labeled as such elsewhere on this page.
A feature checklist makes “build” look cheap. A responsibility checklist tells the truth: every row you build in-house is a system your team staffs, hardens, and maintains indefinitely.
| Responsibility | Build in-house | License EVE |
|---|---|---|
| Deterministic enforcement engine | Design, test & maintain a zero-LLM rules kernel | ✓ Available (pilot) — evaluated in-process, sub-millisecond |
| Regulatory policy packs & updates | Staff policy engineers; track ECOA, FCRA, SR 26-2, HIPAA, EU AI Act as they change | ✓ 27 packs, maintained with version history & effective dates |
| Cryptographic evidence platform | Certificate schema, canonicalization, signing & verification | ✓ Ed25519 certs, RFC 8785 canonicalization, offline verify |
| Key management & rotation | Generate, rotate & revoke keys; keep old evidence verifiable | ✓ Rotation with historical continuity; HSM-aware |
| Replay resistance & at-most-once authorization | Prevent duplicate / replayed approvals under concurrency | ✓ Replay guard + at-most-once authority consumption |
| Exact-payload binding | Bind approval to the precise action; detect post-approval tampering | ✓ Canonical request / response binding enforced |
| Offline, multi-language verifier | Ship & maintain Python + TypeScript verifiers auditors can run | ✓ eve-coreguard (Python) + eve-ai-governance (TS) |
| Tamper-evident audit + Merkle aggregation | Hash-chained log, signed roots, inclusion proofs | ✓ Hash-chained audit bus + signed Merkle roots |
| Fail-closed enforcement service | Stay correct through outages, retries & partial deploys | ✓ Fail-closed by default across the pipeline |
| Deployment surface | Sidecar, API, self-host, HA, multi-region | ✓ Sidecar, API & self-host; appliance in preview |
| Evidence structured for examiner review | Turn logs into evidence structured for examiner and auditor review | ✓ Signed decision records + downloadable evidence packs |
| Multi-year compatibility & retention | Keep certificates verifiable across schema & infra changes for years | ✓ Versioned schema + retention-safe evidence |
“Build in-house” describes work your team owns and maintains indefinitely, not a one-time task. The EVE column reflects capabilities available in controlled, pilot deployments (EVE CoreGuard is pre-GA); items in preview or maturing are labeled as such. SR 26-2 (revised interagency model-risk guidance) applies where relevant — generative and agentic AI sit outside its formal scope and are governed under other risk and compliance frameworks.
Don't take the row on faith — verify the shipping claims directly: sub-millisecond benchmark · offline signature verification · examiner evidence pack · control-plane architecture · readiness & security posture.
Every artifact below is generated from EVE's own code by a reproducible script and carries its commit, generation date, and a SHA-256 self-digest. Download and verify offline — no EVE involvement required.
| Claim | Status | Direct evidence |
|---|---|---|
| Signed decisions, verifiable with a public key | Reproducible vector | signed-decision-test-vector.json — Ed25519 signature + public key; independently verified with the crypto library, not EVE's own verify() |
| Cross-language verification (Python + TypeScript) | SDK-verifiable | Same Ed25519 vector checks under any public-key verifier — eve-coreguard (Python), eve-ai-governance (TS); try it at /verify |
| Signed Merkle roots + inclusion proofs | Reproducible proof | merkle-inclusion-proof.json — inclusion proof verifies against the root; a non-member hash is rejected |
| Key-rotation historical continuity | Reproducible fixture | key-rotation-continuity-fixture.json — a pre-rotation signature still verifies against the old public key after rotating to a new one |
| At-most-once authorization | Tested — 22 pass | governance-test-report.json — captured run of tests/test_execution_token.py |
| Fail-closed enforcement | Tested — 8 pass | governance-test-report.json — captured run of the production-seal fail-closed suite |
| 27 regulated policy packs | Enumerated from source | policy-pack-catalog.json — all 27 packs listed straight from the policies package |
| Sub-millisecond enforcement | Benchmark | /benchmark — latency methodology and results |
Artifacts are demonstration fixtures generated at the current commit; regenerate with python scripts/generate_verification_artifacts.py. Independent third-party penetration and cryptographic review is in progress (see “What's maturing” below). Concurrency and outage/bypass coverage beyond the two suites above runs in CI; downloadable reports for those are being expanded.
A license replaces the need to own and operate everything on the EVE side of the division-of-labor split — and every row in the responsibility matrix above — permanently. That is the real comparison: not EVE's price against a sprint, but against the fully-loaded, multi-year cost of running those functions in-house (see the five-year estimate).
Build the AI workflows that are unique to your business. Licensing EVE means you don't spend the next several years building — and then defending — the governance substrate underneath them. A controlled pilot starts at $37,500 for 60 days, and maps one of your regulated decisions end to end.
A prototype is cheap; a maintained governance substrate is a multi-year operating cost. Adjust the assumptions to your organization — the estimate is deliberately conservative and shows a range, not one dramatic number. For a fuller model, use the ROI calculator.
Estimate only, from your inputs — not a quote. The in-house figure is fully-loaded headcount (build plus ongoing policy, key, reliability and evidence operations) with independent security review, infrastructure and retention, shown as a −15% / +25% band. It excludes opportunity cost and time-to-production risk. The EVE figure is illustrative, based on the published $150,000/yr Enforcement license with the 60-day pilot credited toward year one; your actual quote depends on decision volume and domain.
EVE isn't for everyone, and a page that pretended otherwise wouldn't be worth trusting. Building your own is a reasonable decision when most of the following hold:
A wrong or unlogged decision carries little regulatory, financial or safety consequence, so examiner-grade evidence isn't required.
No regulator, counterparty or auditor will ever need to independently check a past decision against the policy that was in force at the time.
You don't need certificates signed today to still verify after years of key rotations, schema changes and infrastructure migrations.
You have standing policy-engineering, applied-cryptography, platform-reliability and regulatory-maintenance capacity you're willing to point at this indefinitely.
If those are true, build it — you'll likely do it well. EVE earns its price precisely where they're not true: regulated, adversarial, audited decisions whose evidence has to hold up years later.
Build your workflows — the AI products and decisions unique to your business are yours to own. License the governance substrate — the enforcement kernel, evidence plane, policy lifecycle and execution authority are a maintained operation, not a feature you finish. Retain organizational accountability — EVE supplies the technical controls and signed evidence; the compliance decisions and legal classification stay yours.
EVE CoreGuard runs in controlled, pilot deployments today — it is pre-general-availability, and no capability below is claimed as independently validated yet. A vendor that overstates a control an examiner will actually test is a liability, not a moat. So here is the line, drawn plainly.
That honesty is itself part of what you're buying: a governance vendor whose claims survive an examiner's follow-up question.
The open SDKs — eve-coreguard (Python) and eve-ai-governance (TypeScript/Node), plus the separately published eve-proof offline verifier — are a client for submitting decisions and offline verifiers for checking signed certificates; no signing happens client-side. The production governance control plane, the policy-management system, the enterprise regulated policy packs, the execution-authority infrastructure, and managed evidence operations are licensed EVE technology. Adopt the verification surface for free; license the substrate underneath it.
Tell us the decision you'd have to govern — lending, claims, trading, care — and we'll walk you through a signed, offline-verifiable decision from EVE CoreGuard and the full scope you'd otherwise own in-house. A controlled 60-day pilot is $37,500 — or $27,500 at the founding design-partner rate while cohort slots remain — and is credited 100% toward your license.
Capabilities described as EVE's map to components in the EVE control plane; items explicitly labeled “maturing,” “roadmap,” “preview,” or “readiness” are not represented as generally available. Product and company names are trademarks of their respective owners. Pilot pricing current as of 2026 and subject to change.