John M. Willis

A system and method are provided for governing execution of actions in an artificial intelligence (AI)-enabled environment. One or more software agents generate action proposals, which are evaluated by a governance control plane prior to execution. The governance control plane applies deterministic validation logic to determine execution eligibility of each action proposal.

In certain embodiments, evaluation is performed at an execution boundary using a representation of current system… Show more

A system and method are provided for governing execution of actions in an artificial intelligence (AI)-enabled environment. One or more software agents generate action proposals, which are evaluated by a governance control plane prior to execution. The governance control plane applies deterministic validation logic to determine execution eligibility of each action proposal.

In certain embodiments, evaluation is performed at an execution boundary using a representation of current system conditions. The validation process may include applying one or more policy constraints to the action proposal and associated contextual information. Execution of the action proposal is permitted only upon satisfaction of the constraints, and is otherwise prevented.

The system may further support structured action representations, contextual metadata association, and recording of evaluation outcomes to enable audit and replay. The approach separates generation of action proposals from enforcement of execution constraints, enabling controlled and consistent operation of AI-enabled systems across varying domains. Show less

A deterministic governance control-plane architecture for managing execution of actions in automated systems. A governance envelope, bound to a tenant and action identifier, is processed through a strictly ordered, non-reorderable evaluation pipeline. Each evaluation stage produces immutable entries in an append-only ledger, assigned strictly increasing sequence values to enforce a per-action total ordering invariant. Lifecycle state is derived exclusively from the ordered ledger entries… Show more

A deterministic governance control-plane architecture for managing execution of actions in automated systems. A governance envelope, bound to a tenant and action identifier, is processed through a strictly ordered, non-reorderable evaluation pipeline. Each evaluation stage produces immutable entries in an append-only ledger, assigned strictly increasing sequence values to enforce a per-action total ordering invariant. Lifecycle state is derived exclusively from the ordered ledger entries, eliminating reliance on independently mutable status fields. Execution is authorized only when a valid authorization reference is present in the ledger and lifecycle state is re-derived immediately prior to execution. Configuration version identifiers may be recorded to support deterministic replay and lifecycle reconstruction. Show less

Traditional intrusion detection systems operate at the packet or flow level and rely solely on classical statistical or neural network models. Q-PAD introduces a novel architectural approach that models communication traffic at the bitwise conversation level — formally defining a conversation as the sequence of packets exchanged between endpoint pairs (IP address and port combinations).

The system integrates classical sequential modeling with variational quantum circuits and a quantum… Show more

Traditional intrusion detection systems operate at the packet or flow level and rely solely on classical statistical or neural network models. Q-PAD introduces a novel architectural approach that models communication traffic at the bitwise conversation level — formally defining a conversation as the sequence of packets exchanged between endpoint pairs (IP address and port combinations).

The system integrates classical sequential modeling with variational quantum circuits and a quantum recurrent state mechanism that evolves across packet sequences. By encoding structured protocol embeddings into entangled quantum states and fusing quantum measurement outputs with classical representations, Q-PAD models cross-field and cross-packet correlations that are not efficiently representable in comparably sized classical architectures.

The invention contributes a full-stack, deployment-ready hybrid quantum–AI system, including deterministic threshold enforcement, distributed processing, and enterprise-compatible integration pathways. It represents a practical bridge between quantum computing theory and real-world secure infrastructure, advancing the field of hybrid quantum–classical AI for operational cybersecurity applications. Show less

This invention introduces a runtime execution-control framework for neural networks that computes feature relevance during the same forward inference pass and dynamically modifies execution scheduling in real time.

Unlike post-hoc explainability techniques, the system alters the computational graph during inference by suppressing low-relevance computation elements before downstream arithmetic operations are dispatched. This reduces unnecessary arithmetic operations, memory fetches, and… Show more

This invention introduces a runtime execution-control framework for neural networks that computes feature relevance during the same forward inference pass and dynamically modifies execution scheduling in real time.

Unlike post-hoc explainability techniques, the system alters the computational graph during inference by suppressing low-relevance computation elements before downstream arithmetic operations are dispatched. This reduces unnecessary arithmetic operations, memory fetches, and processor energy consumption while preserving predictive performance.

The architecture further enforces lifecycle stability by computing cross-phase relevance drift across foundational training, task-specific training, deployment, and post-deployment monitoring phases. A structured lifecycle state engine stores relevance and drift metrics and selectively authorizes parameter updates for unstable components.

Optional embodiments include hardware-aware scheduling and hybrid classical–quantum execution coordination. Show less

Introduces an ecosystem-level governance control plane that partitions enterprise AI environments into isolated operational domains with defined authority scopes and policy constraints.

Governance is externalized from individual tools into a deterministic control layer that synchronizes execution invariants across domains. The architecture constrains state transitions through validation of authority scope, contextual integrity, and structured contract compliance prior to… Show more

Introduces an ecosystem-level governance control plane that partitions enterprise AI environments into isolated operational domains with defined authority scopes and policy constraints.

Governance is externalized from individual tools into a deterministic control layer that synchronizes execution invariants across domains. The architecture constrains state transitions through validation of authority scope, contextual integrity, and structured contract compliance prior to action.

Enables scalable, state-aware governance of distributed AI-enabled cybersecurity operations while preserving structural isolation and cross-domain integrity. Show less

Describes a deterministic governance architecture for integrating probabilistic AI outputs into operational decision systems under explicit policy constraints.

The invention defines a control-plane model that:

• Externalizes policy enforcement from AI agents
• Implements deterministic control gates around autonomous actions
• Establishes structured execution context and authorization validation
• Governs transitions between automated reasoning and human oversight
•… Show more

Describes a deterministic governance architecture for integrating probabilistic AI outputs into operational decision systems under explicit policy constraints.

The invention defines a control-plane model that:

• Externalizes policy enforcement from AI agents
• Implements deterministic control gates around autonomous actions
• Establishes structured execution context and authorization validation
• Governs transitions between automated reasoning and human oversight
• Preserves traceability, reproducibility, and bounded execution

Designed to ensure that AI-enabled systems remain policy-bound and operationally coherent in enterprise, regulated, and safety-critical environments. Show less

Automated system for security functional requirements generation and structured risk modeling within enterprise SSDLC processes.

Presented at International Computer Security Symposium and SABSA World Congress (2016).