Deterministic engines for AI language, decisions, actions, and oversight in regulated environments.
FERZ's portfolio comprises five pathway engines, a cross-cutting engine, and two governance architectures for regulated AI environments. The pathway engines address distinct governance requirements within an intended architecture for deterministic runtime authorization and evidence. The governance architectures describe how those capabilities compose across recursive and multi-domain oversight.
FERZ is developing runtime authorization infrastructure that makes release conditional on an action-specific verdict and preserves an authorization artifact for independent examination. The portfolio brings together semantic evaluation, constraint evaluation, bounded adaptation, causal modeling, and distributed authority within that architecture.
How the portfolio fits together
The five pathway engines address different parts of the governance problem: meaning and coherence, executable constraints, bounded adaptation, causal modeling, and distributed authority. Their roles differ, and not every engine produces an authorization verdict.
The architecture specifies a common interoperability model and record discipline. LASO(f), DELIA, and Constitutional Blockchain are specified to produce authorization verdicts. CausaCore and the FERZ Behavioral Engine produce causal and adaptive outputs that can inform an authorization evaluation. BiasConstraint addresses bias and fairness requirements across applicable pathways.
Composition is specified through linked evidence records, with integration scope defined for the selected workflow. The intended result is an audit trail that preserves each participating engine's contribution and its relationship to the authorization verdict.
Implementation status. The reviewed implementation provides content evaluation, integrity-checked records, and in-process verification, with enforcement mode recorded. Remaining development includes action release, keyed integrity, signing, and independent reconstruction from an exported evidence package. Scope and sequencing are subject to financing and prioritization. Pilots are scoped accordingly.
The product portfolio
Multiple entry points, organized by architectural role. Five pathway engines address distinct governance requirements. BiasConstraint adds bias and fairness constraints across applicable pathways. STRATA-G and VECTARA(f) describe how governance capabilities compose into recursive and multi-domain oversight.
Five pathway engines, each designed around a distinct architectural approach to deterministic governance. The descriptions below state their intended roles; status is identified for each engine.
LASO(f)
Deterministic governance for language, action, and meaning
PathwaySemantic Determinism, for meaning and coherence
Intended roleEvaluates AI language and action outputs against codified semantic standards for meaning and coherence. The specified output is ALLOW, DENY, or ABSTAIN with an authorization artifact. Action release is governed by the runtime authorization boundary.
Risk addressedPolicy-nonconforming language, unsafe instructions, and semantic mismatches between an output and the standards it must satisfy.
StatusAvailable for pilot and implementation discussions.
Deterministic Executive Layer for Interpretable Alignment
PathwayConstraint Determinism, for bounded executable constraints
Intended roleEvaluates proposed AI outputs and actions against codified constraints. The specified output is ALLOW, DENY, or ABSTAIN with an authorization artifact. In the intended architecture, the runtime authorization boundary conditions release on the action-specific verdict.
Risk addressedOutputs and actions that violate encoded constraints, and missing evidence of the basis for an authorization verdict.
StatusAvailable for pilot and implementation discussions.
PathwayAdaptive Determinism, for deterministic bounded adaptation
Intended roleAdapts AI output to operational context within codified ceilings. The specified output pairs adapted content with an explainable rationale and a record of the trajectory state, modifiers applied, and ceiling in force. It does not produce an authorization verdict.
Risk addressedAdaptation that responds to context without remaining within declared bounds, and static responses that do not account for relevant operational context.
PathwayCausal Determinism, for deterministic causal modeling across multi-tier and cross-domain systems
Intended roleModels causal structure across hierarchical tiers and domain boundaries, with an explainable causal pathway and a record traceable from input to output. It supplies causal output that can inform analysis or authorization; it does not produce an authorization verdict.
Risk addressedOpaque causal reasoning, analysis confined to a single domain, and missing traceability between inputs and causal outputs.
StatusAvailable for pilot and implementation discussions.
PathwayConsensus Determinism, for distributed authority over governing rules, decision records, and halt operations.
Intended roleSpecifies distributed authority over governing rules, decision records, and halt operations across multiple validators. Its specified output is a consensus verdict and a record of the distributed authority decision.
Risk addressedUnilateral control over governing rules, records, or halt decisions where authority is required to be distributed.
StatusArchitecture specified; implementation roadmap in development.
A cross-cutting engine for applying bias and fairness requirements within the applicable governance pathways.
BiasConstraint Engine
Constraint enforcement for bias and fairness
Intended roleApplies codified bias and fairness requirements as constraints within applicable governance pathways, with evidence of the checks performed and their results specified as part of the record.
Risk addressedBias and fairness requirements that remain untested or lack a traceable record of evaluation.
STRATA-G and VECTARA(f) are published governance architectures describing how capabilities compose into recursive and multi-domain oversight. They are architecture specifications, not deployable engines or additional determinism pathways. FERZ's intended product architecture assigns operational roles to the pathway engines within these broader arrangements.
STRATA-G
Published recursive governance architecture
Intended roleDescribes recursive composition of governance capabilities, including constraint enforcement, escalation, and oversight. It provides an architectural arrangement for assigning and connecting those responsibilities.
Risk addressedUnclear escalation responsibilities, fragmented oversight, and failure to intervene when authorized human resolution is required.
Intended roleDescribes how governance capabilities compose across multiple domains when one system contributes to decisions in several domains at once.
Risk addressedDecisions that satisfy local requirements while creating unmanaged dependencies or risk across domains.
No. The engines are offered for individual licensing, with implementation scope defined for the engagement. A deployment can begin with one pathway and add others as its requirements develop.
How do the engines work together?
The architecture specifies a common interoperability model and record discipline. Records are specified to link through common fields so that each engine's contribution can be examined within the composed workflow. Integration scope is defined for the selected engines and use case.
Can FERZ integrate with our existing AI systems?
FERZ's intended architecture places authorization at the boundary between a proposed action and its governed effect, alongside the model, application, and identity and access layers. Integration depends on a release path that the agent cannot bypass. The applicable placement and integration scope are established for the environment.
Can we pilot before a full deployment?
Pilot and implementation discussions are available for LASO(f), DELIA, and CausaCore. A pilot can focus on one engine and a defined workflow, with scope matched to the available implementation and agreed evaluation criteria.