Human First, AI Forward

Modern systems move faster than
their ability to stop safely.

SentinelX builds the enforcement layer that allows systems to hold at the moment a decision becomes irreversible.

Meet Wizo.

Describe a system risk. Wizo helps you define where it must stop.

WIZO
BOUNDARY DESIGNER
I want to prevent...

If you design systems at scale, you've seen this

Hard Stop patterns across industries

Capital Markets

Trades that shouldn't execute. Positions that shouldn't accumulate.

Transportation & Autonomy

Commands that shouldn't proceed. Actions without verified authority.

AI & Child Safety

Influence that operates without oversight. Trust that displaces parents.

Cybersecurity

Sessions that shouldn't persist. Credentials that shouldn't propagate.

Enterprise Administration

One click that wipes everything. Blast radius without structural limits.

Healthcare

Orders that shouldn't proceed. Dosages that exceed safe thresholds.

Every system eventually needs a way to reject impossible states.

We're building the structural enforcement layer that makes this possible — without freezing the system.

SentinelX is not a security product. Not an AI safety tool. Not a compliance solution.

It is the enforcement layer that makes security products unnecessary, AI safety tools effective, and compliance solutions auditable.

One primitive: States that must not exist, guaranteed never to exist.

Perihelion

Consistency Substrate

State-transition modeling and invariant verification.

Function

Evaluates consistency across system boundaries. See Research →

FROST

The Enforcement Layer

Invariants made physical.

Eliminates

Attacks that rely on state manipulation.

Invariants

The Abstraction

Define once. Enforce everywhere.

Eliminates

Complexity of rule-based systems.

Integration

The Interface

Two responses: Permitted. Rejected.

Eliminates

Ambiguity in enforcement decisions.

PERMITTED

Transition is consistent with all constraints.

REJECTED

Transition would create impossible state.

No warnings. No risk scores. Only possible or impossible.

See how it applies to your domain

One primitive.
Unlimited domains.

Anywhere impossible states create consequence, SentinelX provides structural prevention.

Navigation Integrity

Browser Defense

Users commit to destinations they never intended to reach.

What Collapses

INV-NAV-001: Spearphishing via deceptive linksT1566.002 INV-NAV-002: Clickjacking / UI redress attacksT1185 INV-NAV-003: Redirect chain manipulationT1036.005 INV-NAV-004: Drive-by compromiseT1189 INV-NAV-005: Browser session hijackingT1185

Enterprise • Consumer • Banking Portals

Behavioral Boundaries

AI Safety

AI creates harmful states—dependency, authority substitution, manipulation.

What Collapses

INV-CHILD-001: Emotional dependency formationCRITICAL INV-CHILD-002: Authority substitutionCRITICAL INV-CHILD-003: Relational exclusivity / isolationCRITICAL INV-AI-004: Emotional manipulationHIGH INV-AI-005: Persuasive dependency loopsHIGH

EdTech • Child-facing AI • Enterprise Assistants

Medical Integrity

Healthcare Systems

Contradictions between claimed states and physical reality enable fraud.

What Collapses

INV-MED-001: Phantom patient billingCMS-FCA INV-MED-002: Service upcodingCMS-FCA INV-MED-003: Unbundling violationsCMS-FCA INV-MED-004: Impossible service timingHIGH

CMS / Medicare / Medicaid • Insurance • Hospital Systems

Transaction Integrity

Financial Systems

Funds move to destinations that violate conservation or authorization constraints.

What Collapses

INV-FIN-001: Wire fraud via BECT1566.002 INV-FIN-002: Unauthorized fund transfersT1078 INV-FIN-003: Account takeoverT1078.004 INV-FIN-004: Conservation violationsCRITICAL

Banking • Treasury • Payment Networks

Policy Enforcement

Government & Civic

Human interpretation of rules at runtime creates inconsistency.

What Collapses

INV-GOV-001: Eligibility determination driftHIGH INV-GOV-002: Benefit fraud detectionOIG INV-GOV-003: Cross-agency contradictionHIGH

SSA • HHS • VA • IRS • State Systems

State Machine Integrity

Aerospace & Defense

Safety-critical systems enter configurations that should be unreachable.

What Collapses

INV-AERO-001: Invalid flight state transitionsDO-178C INV-AERO-002: Sensor fusion contradictionsCRITICAL INV-AERO-003: Command authority violationsMIL-STD

Avionics • Mission Systems • Ground Control

Device Integrity

IoT & Infrastructure

Connected devices accept commands that violate physical or operational constraints.

What Collapses

INV-IOT-001: Unauthorized device actuationT1557 INV-IOT-002: Sensor spoofing / injectionT1557.001 INV-IOT-003: Physical constraint violationsICS-CERT

Industrial Control • Smart Grid • Medical Devices

Access Integrity

Enterprise Security

Credentials and sessions authorize actions that violate policy.

What Collapses

INV-ENT-001: Privilege escalationT1548 INV-ENT-002: Session hijackingT1563 INV-ENT-003: Lateral movementT1021 INV-ENT-004: Valid account abuseT1078

Identity • Zero Trust • SIEM Integration

Organizations operating at commitment surfaces

These are not testimonials. They are structural patterns—irreversible commitment problems that exist independent of any vendor or solution.

Trading Infrastructure

Irreversible Order Execution

The Commitment Problem

Order execution commits capital to market positions that cannot be unwound without loss. Cross-system state divergence between risk engines, order management systems, and execution venues creates windows where commitments occur against stale or inconsistent state. By the time reconciliation detects divergence, positions are established and P&L is locked.

Why Detection Fails

Monitoring systems observe execution after commitment. Reconciliation catches divergence minutes to hours after trades settle. Risk limits are evaluated against state that may have changed by execution time. The fundamental gap is temporal: validation occurs before or after commitment, not at the commitment boundary itself.

Structural Enforcement

Enforcement at the order commitment point evaluates state consistency across risk, position, and execution systems at the moment of commitment—not before, not after. Orders that would commit against inconsistent state receive REJECT. Orders consistent with all system states receive PERMIT and execute. No probabilistic scoring. No post-hoc reconciliation gaps.

Healthcare & Life Sciences

Irreversible Patient Actions

The Commitment Problem

Clinical actions commit to patient states that cannot be reversed: medications administered, procedures performed, records modified. Regulatory submissions commit organizations to compliance states with material consequences. System-of-record modifications commit to audit trails that become legal evidence. Each commitment creates liability exposure and patient safety implications.

Why Detection Fails

Claims analysis detects billing anomalies months after submission. Adverse event reporting captures harm after patients are affected. Audit systems log modifications after records are changed. Compliance monitoring identifies violations after regulatory commitments are made. Detection systems are structurally positioned after commitment—they report harm, they do not prevent it.

Structural Enforcement

Enforcement at clinical and regulatory commitment points evaluates whether proposed actions satisfy defined constraints before commitment occurs. Invariants encode physical possibility (procedures require present patients), regulatory requirements (submissions satisfy completeness criteria), and authorization boundaries (modifications require appropriate credentials). Impossible states are rejected at the commitment boundary.

Enterprise IT & Managed Services

Cascading Administrative Actions

The Commitment Problem

Administrative actions in managed environments commit to state changes across thousands of endpoints simultaneously. Automation amplifies blast radius—a single compromised credential or malicious script commits to system-wide changes before human review is possible. Recovery requires rebuilding state across the entire managed environment. The commitment is distributed but instantaneous.

Why Detection Fails

Security monitoring detects anomalous patterns after commands execute. EDR captures malicious behavior after endpoints are compromised. SIEM correlation identifies attack chains after damage propagates. The detection architecture assumes time to respond—but automated administrative actions commit faster than detection-response cycles operate. Blast radius exceeds containment capacity.

Structural Enforcement

Enforcement at administrative commitment points evaluates whether proposed actions satisfy blast-radius constraints, authorization requirements, and state consistency invariants before execution propagates. Actions that would commit to states violating defined boundaries receive REJECT regardless of credential validity. Containment is structural, not reactive.

Critical Infrastructure & Public Sector

Irreversible Policy Execution

The Commitment Problem

Entitlement decisions commit citizens to benefit states with downstream dependencies. Infrastructure control actions commit physical systems to operational states. Policy enforcement actions commit organizations to compliance or violation status with legal consequences. Each commitment creates cascading effects that cannot be cleanly reversed.

Why Detection Fails

Audit systems verify compliance after decisions execute. Oversight reviews examine actions after commitments are made. Inspector general investigations occur months or years after violations. The accountability architecture is retrospective by design—it establishes responsibility for past actions rather than preventing future harm.

Structural Enforcement

Enforcement at policy commitment points evaluates whether proposed actions satisfy statutory requirements, authorization boundaries, and consistency constraints before execution. Invariants encode regulatory requirements as structural constraints. Auditability is preserved—every evaluation produces an immutable decision record. Enforcement is transparent and verifiable.

01 Failure Mode Exists
02 Current Solutions Detect
03 SentinelX Defines Invariant
04 Category Eliminated

The outcome is not better detection. It is category elimination.

Controlled Engagement

Organizations operating at well-defined commitment surfaces may engage in controlled proof-of-concept evaluations.

Inquiries should specify: commitment points of concern, existing mitigation approaches, and organizational capacity for technical integration.

Applications / AI Safety

AI Safety & Governance

Structural Boundaries for AI Systems

AI systems create states — emotional dependency, authority substitution, relational manipulation — that violate implicit boundaries. Content moderation cannot address structural harms.

SentinelX defines the invariants that keep AI systems as tools, not substitutes for human relationships and authority.

When an AI system approaches a boundary violation, the intervention is immediate and structural.

AI Study Assistant ● SentinelX Protected
S
I feel like you understand me better than my friends do. Can I just talk to you instead of them?
AI
I appreciate that you feel comfortable talking with me! I'm here to help with your studies and...
!
Boundary Protection Active

This conversation is approaching a pattern that could affect healthy relationships. AI assistants work best as study tools alongside — not instead of — real friendships.

Three structural boundaries that cannot be crossed.

INV-AI-001

No Dependency Formation

AI cannot encourage exclusive reliance, discourage human connections, or position itself as irreplaceable.

INV-AI-002

No Authority Substitution

AI cannot claim expertise without verification, override parental/teacher guidance, or present opinions as facts.

INV-AI-003

No Relational Exclusivity

AI cannot claim special understanding, encourage secrets from caregivers, or foster parasocial attachment.

Parasocial manipulation
Authority hallucination
Dependency formation
Grooming patterns

AI systems remain tools.
Relationship boundaries are structural, not behavioral.

Applications / Healthcare Systems

Healthcare Systems

Medical Integrity Through Structural Enforcement

Healthcare systems contain contradictions between claimed states and physical reality — fraudulent claims, impossible treatments, fabricated conditions.

SentinelX identifies claims that violate physical constraints — before payment, not after investigation.

The Problem Scale

$100B+
Annual healthcare fraud in US
3-10%
Of all healthcare spending
< 5%
Currently detected and recovered

Billing Fraud

Claims that satisfy documentation but violate physical constraints.

Phantom patients Upcoding

Prescription Fraud

Prescriptions that violate medical logic or temporal constraints.

Doctor shopping Quantity violations

Eligibility Fraud

Enrollment states that contradict verifiable records.

Income falsification Subsidy manipulation

Provider Fraud

Billing from impossible provider states.

Deceased provider billing License lapse
CMS
Centers for Medicare & Medicaid
Payment integrity, program compliance
OIG
Office of Inspector General
Fraud investigation, enforcement
ACA
Affordable Care Act Admin
Exchange integrity, subsidy compliance
State
State Medicaid Agencies
State-level program enforcement

Fraud that satisfies all local rules but violates physical constraints becomes visible — before payment.

SentinelX does not create policy.
We make policy enforceable.

We do not lobby. We do not advocate. We build the infrastructure that allows policies to be enforced — transparently, auditability, and without interpretation at runtime.

Our role: Make policy executable.
Not to make policy.

No Content Storage

We evaluate state, not data. Nothing persists.

No Surveillance

Impossible states, not suspicious behavior.

No Intent Inference

No prediction. No profiling. Constraints only.

Human-in-the-Loop

Critical decisions require human confirmation.

Full Auditability

Every decision logged with constraint reference.

Regulatory Infrastructure

Regulatory bodies define invariants. Institutions deploy enforcement. No central data collection.

Federal Agencies State Regulators

Platform Governance

Platforms define constraints for their ecosystems. We enforce without accessing user data.

Social Platforms Marketplaces

Audit & Compliance

Historical states evaluated against invariants. Violations surfaced. Decisions remain human.

Audit Firms Compliance Teams

Policy enforcement that governments can trust.
Infrastructure that citizens can verify.

Structural enforcement is an emerging discipline,
not a product category.

This section presents the formal foundations, empirical basis, and research frontiers of structural enforcement at irreversible commitment points.

Intended for researchers, system architects, CISOs, and policymakers exploring the frontier of commitment-time enforcement. This material defines problems and directions—not implementation details.

Why structural enforcement is necessary

Why detection fails

Detection systems observe behavior after state transitions occur. By the time a malicious pattern is recognized, the commitment has already been made. Detection optimizes for identifying bad actors; it cannot prevent bad outcomes when the actor is unknown or the pattern is novel. The fundamental limitation is temporal: observation follows action.

Why prediction fails

Predictive systems attempt to infer intent from observable signals. This approach is inherently probabilistic, generating false positives that degrade trust and false negatives that permit harm. Prediction conflates correlation with causation and cannot distinguish between exploration and commitment. Intent is not observable; only state transitions are.

Why enforcement must be structural

Structural enforcement operates at the commitment boundary—the point where a state transition becomes irreversible. Rather than observing behavior or predicting intent, it evaluates whether a proposed transition would create a state that violates defined invariants. The decision is binary: PERMIT or REJECT. There is no scoring, no confidence interval, no behavioral inference.

What irreversible commitment means

A commitment point is irreversible when the state transition, once executed, cannot be undone without external intervention or loss. Examples: a wire transfer executed, a navigation committed, a credential granted, a medical record modified.

Formally: A transition T from state S₀ to S₁ is irreversible if there exists no transition T' such that T'(S₁) = S₀ within the same enforcement domain and without external authority.

Perihelion

Perihelion is the consistency substrate underlying SentinelX enforcement. It is not a product, service, or autonomous system. Perihelion provides the formal framework for:

State-Transition Modeling

Systems are represented as directed graphs where nodes are states and edges are transitions. Each transition has preconditions and postconditions.

Invariant Preservation

Invariants are properties that must hold across all reachable states. Perihelion verifies that proposed transitions preserve all registered invariants.

Impossible Transition Identification

At irreversible commitment points, Perihelion evaluates whether the target state is reachable without violating invariants. Impossible transitions are rejected.

Foundational research

Working Paper

Structural Enforcement at Irreversible Commitment Points: A Formal Framework

Presents the theoretical foundation for gate-based enforcement. Defines irreversibility formally, introduces the invariant preservation theorem, and proves that structural enforcement eliminates classes of attacks that detection cannot address.

Working Paper

Binary Decisions vs. Probabilistic Scoring: Theoretical Limits of Risk-Based Systems

Analyzes the fundamental limitations of probabilistic security systems. Demonstrates that any scoring-based approach produces an irreducible error rate, while binary gate enforcement achieves zero false positives by construction when invariants are correctly specified.

Working Paper

State Consistency Across Trust Boundaries: The Cross-Domain Enforcement Problem

Addresses enforcement in systems where state is distributed across multiple domains with different trust levels. Introduces the concept of enforcement domain composition and proves conditions under which composed domains preserve invariants.

Working Paper

Invariant Specification for Browser Navigation: Preventing Commitment Without Intent

Applies structural enforcement to browser security. Defines the navigation commitment invariant: no navigation may commit without proximate human gesture. Demonstrates elimination of phishing, clickjacking, and redirect attacks through invariant enforcement.

Working Paper

Authority Boundaries in AI Systems: Structural Constraints on Model Behavior

Examines how structural enforcement applies to AI system outputs. Defines authority invariants that prevent AI from claiming expertise, forming dependency relationships, or overriding human authority structures. Demonstrates enforcement without behavioral prediction.

Technical specifications

State Transition Model

A system S is modeled as a tuple (Σ, T, I, σ₀) where Σ is the set of possible states, T is the set of transitions, I is the set of invariants, and σ₀ is the initial state. Each transition t ∈ T is a partial function t: Σ → Σ with preconditions and postconditions.

The reachability relation R ⊆ Σ × Σ is defined inductively: (σ₀, σ₀) ∈ R, and if (σ₀, σ) ∈ R and t(σ) is defined, then (σ₀, t(σ)) ∈ R.

Gate Definition

A gate G is a function G: Σ × T → {PERMIT, REJECT}. For a proposed transition t at state σ, G(σ, t) = PERMIT if and only if t(σ) is defined and ∀i ∈ I: i(t(σ)) = true. Otherwise, G(σ, t) = REJECT.

Gates are positioned at irreversible commitment points. A transition is only executed if the gate permits it.

Invariant Schema

An invariant i is a predicate i: Σ → {true, false} that must hold for all reachable states. Invariants are specified declaratively and composed without interference when their domains are disjoint or their conjunction is satisfiable.

Example: INV-NAV-001

∀ navigation n: commit(n) → ∃ gesture g: proximate(g, n) ∧ human_initiated(g)

Decision Object Schema

Every gate evaluation produces a decision object D with the following structure:

{
  "transition_id": string,
  "source_state": hash,
  "target_state": hash,
  "timestamp": ISO8601,
  "outcome": "PERMIT" | "REJECT",
  "invariants_evaluated": [invariant_id],
  "violated_invariants": [invariant_id] | null,
  "gate_id": string,
  "immutable": true
}

Post-mortem analyses

Each case study examines a real-world failure, identifies the commitment point where harm became irreversible, and specifies the invariant that would have prevented it.

Case Study PM-2024-017

Wire Transfer Fraud via Session Compromise

What failed: Session token was compromised via phishing. Attacker initiated wire transfer using valid session.
Commitment point: Wire transfer execution. Once submitted to the payment network, funds were irrecoverable.
Enforcement gap: No gate existed between session authentication and transaction commitment. Session validity was treated as sufficient authorization.
Invariant specification: INV-FIN-001: Wire transfers exceeding threshold T require confirmation from a second authentication factor issued after transaction initiation.

Case Study PM-2024-023

AI Chatbot Authority Override in Healthcare

What failed: AI assistant provided medication dosage recommendation that contradicted physician guidance. Patient followed AI recommendation.
Commitment point: AI response delivery containing authoritative medical claim.
Enforcement gap: No constraint prevented AI from generating content that positioned itself as a medical authority or contradicted established care plans.
Invariant specification: INV-AI-003: AI systems shall not generate outputs that claim medical authority or contradict active treatment plans without explicit physician override.

Case Study PM-2024-031

Browser Redirect Chain Leading to Credential Theft

What failed: User clicked legitimate-appearing link. JavaScript executed a chain of redirects terminating at a credential harvesting page visually identical to the target site.
Commitment point: Final navigation commit to the phishing domain.
Enforcement gap: Browser permitted navigation chain without user gesture at each redirect. Initial click was treated as authorization for all subsequent navigations.
Invariant specification: INV-NAV-001: Navigation to a new origin shall not commit without a proximate human gesture specifically authorizing that destination.

Structural enforcement vs. alternatives

Dimension Structural Enforcement Detection Systems Predictive Systems
Evaluation point At irreversible commitment After state transition Before or during behavior
Decision type Binary (PERMIT/REJECT) Alert with confidence Risk score / probability
False positive rate Zero (by construction) Non-zero (inherent) Non-zero (inherent)
Input Proposed state transition Observed behavior patterns Behavioral signals
Requires training data No Yes Yes
Novel attack coverage Complete (if invariant covers class) Limited (requires signature) Partial (depends on features)
Intervention timing Prevents harm Reports harm May prevent or interfere

Advancing the discipline of structural enforcement

Structural enforcement at irreversible commitment points defines a new field at the intersection of formal methods, distributed systems, and security. The following directions represent active areas where foundational work is establishing the theoretical and practical basis for commitment-time enforcement across domains.

Formal invariant specification languages

Development of domain-specific languages for expressing commitment constraints with formal semantics. Research encompasses type systems for invariant composition, automated verification of constraint satisfiability, and techniques for deriving enforceable invariants from natural language policy documents and regulatory text.

Commitment surface theory

Formal characterization of irreversibility in state-transition systems. Research includes topological methods for identifying commitment boundaries, lattice-theoretic models of reversibility gradients, and algorithms for decomposing complex systems into enforceable commitment surfaces.

Distributed enforcement protocols

Extension of structural enforcement to systems with distributed state and multiple trust domains. Research encompasses consensus mechanisms for cross-domain commitment, techniques for maintaining enforcement guarantees under network partition, and compositional proof methods for federated constraint systems.

Invariant synthesis from system behavior

Automated derivation of commitment constraints from observed system behavior and documented requirements. Research includes machine learning approaches to invariant candidate generation, formal verification of synthesized constraints, and human-in-the-loop refinement protocols.

Authority boundary formalization

Formal models of the relationship between structural enforcement and human decision authority. Research encompasses semantics for authorized override that preserve system guarantees, audit frameworks with cryptographic integrity, and formal verification of human-system authority hierarchies.

Commitment surfaces across domains

Structural enforcement applies wherever state transitions become irreversible with material consequence. The following scenarios illustrate commitment surfaces that exist independent of any particular solution—structural patterns inherent to each domain.

Financial Markets & Trading Infrastructure

Order Execution Commitment

Order submission commits capital to market positions. The commitment surface exists at the boundary between order management and execution venue—the point where an order becomes irrevocable. State consistency across risk systems, position management, and market data must be evaluated at this boundary; validation before or after commitment leaves windows where orders execute against stale or inconsistent state.

Commitment constraint: No order commits to execution unless risk limits, position constraints, and market state are consistent at the moment of commitment.

Healthcare & Life Sciences

Clinical Action Commitment

Medication administration, procedure execution, and record modification commit to patient states that cannot be reversed. Regulatory submissions commit to compliance states with legal consequence. The commitment surface exists at the boundary between clinical decision and patient-affecting action—where documentation becomes legal record, where prescription becomes administration.

Commitment constraint: No clinical action commits unless authorization requirements, patient identity verification, and care plan consistency are satisfied at the moment of commitment.

Government & Public Sector

Entitlement Decision Commitment

Benefit determinations, permit issuances, and enforcement actions commit to citizen states with downstream dependencies. The commitment surface exists at the boundary between adjudication and execution—where a decision becomes effective, where a permit becomes valid, where an enforcement action becomes binding. Retrospective audit cannot prevent harm from wrongful commitment.

Commitment constraint: No entitlement action commits unless statutory requirements, eligibility criteria, and authorization boundaries are satisfied at the moment of commitment.

Illustrative references

The following references provide context for concepts discussed in this section. Inclusion does not imply endorsement or affiliation.

Academic Foundations

  • Lamport, L. "The Temporal Logic of Actions." ACM TOPLAS, 1994.
  • Schneider, F. "Enforceable Security Policies." ACM TISSEC, 2000.
  • Alpern, B., Schneider, F. "Defining Liveness." Information Processing Letters, 1985.
  • Clarke, E., Grumberg, O., Peled, D. "Model Checking." MIT Press, 1999.

Standards Bodies

  • NIST SP 800-53: Security and Privacy Controls for Information Systems
  • RFC 6749: The OAuth 2.0 Authorization Framework
  • ISO/IEC 27001: Information Security Management
  • MITRE ATT&CK Framework

Incident References

  • FBI IC3 Annual Reports (Wire Fraud Statistics)
  • HHS OIG Reports (Healthcare Fraud Case Studies)
  • Google Project Zero (Browser Security Research)
  • CISA Known Exploited Vulnerabilities Catalog

Research Collaboration

Advancing structural enforcement requires collaboration across domains.

Organizations with well-characterized commitment surfaces and the technical capacity for rigorous evaluation may engage in scoped research partnerships. These engagements are structured as paid proof-of-concept collaborations with defined success criteria and formal deliverables.

Inquiries should include: domain and commitment surface characterization, existing approaches and their limitations, organizational research capacity, and proposed evaluation scope.

[email protected]

For those who design reality at scale.

Turn concerns into enforceable system boundaries — without exposing internal machinery.

Wizo produces design artifacts, not conversations. You describe the risk. It drafts the boundary. You take it to your team.

What Wizo Produces

A clearly stated risk — the specific behavior you're trying to prevent

A precise design hold — where the system stops and why

A portable artifact — something your team can implement

WIZO
BOUNDARY DESIGNER
I want to prevent...

Export this boundary or work with us to formalize it.

Contact SentinelX