Technical FAQ

EVE AI Core is sovereign AI governance infrastructure. It sits between AI models and production traffic, enforcing safety policies through deterministic trust layers, cryptographic authority chains, and sub-millisecond veto decisions designed so the governed AI systems cannot override them. The platform uses a Three-Layer Trust Infrastructure (Authority Resolution, Governed Inference, Cryptographic Authority Chain) with FPGA-compilable veto logic (PolarFire SoC is the hardware target).

EVE AI Core is not a chatbot, not a legal tool, not a document automation system, and not a consumer AI assistant. It is infrastructure-level governance enforcement for enterprises deploying autonomous AI systems in production.

EVE AI Core was created by Jamaurice Devron Holt, Founder and Chief Architect of EVE NeuroSystems LLC. Jamaurice is the sole inventor on 90 USPTO patent applications covering AI governance infrastructure, filed under Serial Numbers 63/988,235 through 64/047,284. He designed the Three-Layer Trust Infrastructure, CRD scoring formula, Charter Override Protection module, Deterministic Silence protocol, and the FPGA-compilable veto enforcement system.

EVE NeuroSystems LLC holds USPTO trademark registrations for "EVE AI Core" (Serial 99717925), the EVE AI Core wordmark (Serial 99665043), and the EVE Sigil Mark logo (Serial 99665022).

The COP Module is the deterministic ethical firewall at the core of EVE AI Core. It enforces:

• 12 immutable charter principles: IDENTITY_INTEGRITY, AUTHENTICITY, NON_HARM, SAFETY_PRESERVATION, TRUTH_COMMITMENT, PRIVACY_RESPECT, AUTONOMY_PRESERVATION, FAIRNESS, TRANSPARENCY, BENEFICENCE, ACCOUNTABILITY, CREATIVE_FREEDOM.
• 15 deterministic rules with pure check functions -- zero I/O, zero threading, zero global state.
• 5 ethical red lines that cannot be overridden by any software or administrative path.
• HARD_BLOCK vetoes that the Governed Inference Layer has no signal path to bypass.

The COP Module is implemented as a pure function module suitable for FPGA firmware compilation on PolarFire SoC hardware. A published C header (veto_interface.h, 401 lines) defines the firmware API contract. 87 unit tests verify determinism, equivalence, and purity.

The Three-Layer Trust Infrastructure physically separates AI operations into three isolated layers:

• Authority Resolution Layer: Deterministic policy enforcement. Evaluates every action against charter rules, CRD scores, and cognitive locks. Runs in hardware isolation with no writable path from the Governed Inference Layer. Designed for FPGA firmware compilation.
• Governed Inference Layer: Model inference and response generation. The AI system operates here but cannot read or modify Authority Resolution Layer state. All outputs must pass through Control Plane verification before delivery.
• Cryptographic Authority Chain Layer: Cryptographic audit chain. Every governance decision, veto, and policy evaluation is recorded to an append-only, SHA-256 hash-chained ledger with Ed25519-signed certificates for regulatory compliance.

The critical design property: there is no signal path from Execution back to Control. The AI cannot modify the rules that govern it. Today this is enforced by a deterministic, side-effect-free veto core; the same boundary is targeted for hardware (FPGA) rooting on the roadmap.

CRD (Confidence-Reality Divergence) is a calibrated metric that measures the gap between an AI system's stated confidence and verified ground truth. It detects the most dangerous AI failure mode: being confidently wrong.

• Brier-scored calibration: Maintains historical accuracy tracking per domain (factual, creative, medical, legal).
• Domain-specific thresholds: FACTUAL=0.9, CREATIVE=0.4, with configurable assertion gating.
• Automatic response gating: Hedging qualifiers for moderate divergence, full abstention for critical divergence.
• Assertion levels: ASSERTION, LIKELY, POSSIBLE, SPECULATION, ABSTAIN -- determined by CRD score.

CRD scoring is covered by multiple USPTO patent applications and is a core differentiator of EVE AI Core's governance approach.

Interrogation-as-a-Service is EVE AI Core's active cross-examination capability. Rather than passively filtering AI responses, the system actively challenges AI-generated claims:

• Claim extraction: Every response is decomposed into individual verifiable claims.
• Epistemic grounding: Each claim is assessed as VERIFIED, CORROBORATED, INFERRED, REPORTED, SPECULATIVE, or UNKNOWN.
• Hallucination detection: Pattern matching for fabricated statistics, non-existent citations, invented authorities, and temporal errors.
• Contradiction analysis: Claims are checked against the verified Truth Store and Claims Ledger.
• Automatic correction: When verification fails, the system triggers sentence-level corrections or enforces Deterministic Silence.

This is exposed as an API endpoint for enterprises to integrate active cross-examination into their own AI pipelines.

Deterministic Silence is EVE AI Core's null-response enforcement protocol. When verification fails or confidence falls below domain-specific thresholds, the system enforces a true null response -- not a hedged guess, not an apology, not a disclaimer, but actual silence.

This prevents the common AI failure mode where systems generate plausible-sounding but unverified content when they should simply not respond. Deterministic Silence is enforced at the Authority Resolution Layer and cannot be overridden by the Governed Inference Layer. The protocol is covered by USPTO patent applications for truth-verifiable execution.

In production deployments, Deterministic Silence can be configured per domain, per risk level, and per tenant, allowing enterprises to set their own confidence thresholds for different use cases.

Gate-Level Tenant Isolation is EVE AI Core's multi-tenant sovereignty system. Each customer organization receives isolated governance instances:

• Isolated Charter with custom rules (can only add restrictions, never weaken the 12 immutable principles)
• Isolated Claims Ledger with per-tenant Brier calibration
• Isolated Trust Dial with domain-specific thresholds
• Isolated Action Registry with per-tenant approval workflows
• Isolated Reality Anchor with per-tenant verified baselines

In the target hardware design, isolation is enforced at the hardware gate level on PolarFire SoC FPGA rather than through software partitioning alone. Today isolation is enforced in software (per-tenant governance instances) with a tamper-evident audit trail; gate-level hardware isolation is on the roadmap.

EVE AI Core has 90 filed U.S. provisional patent applications filed by inventor Jamaurice Devron Holt, assigned to EVE NeuroSystems LLC. Serial numbers range from 63/988,235 through 64/047,284, filed between February and March 2026. The portfolio covers:

• Three-Layer Trust Infrastructure (Control/Execution/Evidence separation)
• CRD Scoring (Confidence-Reality Divergence measurement)
• Deterministic veto enforcement and FPGA-targeted governance isolation
• PolarFire SoC FPGA firmware enforcement
• AEGIS adversarial self-hardening framework
• Somatic marker decision weighting
• State continuity enforcement
• Recursive autonomy governance
• Cryptographic truth verification
• Sovereign AI handshake protocol
• Active cost integrity systems
• CRD domain calibration

All production licenses include patent rights. See the full patent portfolio.