Runtime Governance for High-Risk AI Systems

When AI Acts,
Governance
Must Act With It.

Make AI Governable.

Piscys is developing an independent control layer between AI
systems and consequential enterprise action, connecting evidence, proportionate safeguards, and accountable decision authority.

Initial Focus: Financial Services and Healthcare.

Advisory Services Available Now. 

Platform in Development.

01 / Control Evidence

From Governance Intent to Control Evidence

When an AI system proposes a consequential action, what supports it, when should safeguards intervene, and who is authorized to decide? Piscys is being designed to connect those questions to reviewable control evidence.

01

Exposure Minutes

A proposed measure from substantiated material-risk onset to verified, accountable containment. Unknown onset or unverified containment must remain visible.

02

AI Trust Metrics

Intended measures of intervention timing, override patterns, control health, and supporting-evidence completeness.

03

Assurance Dashboard

A proposed shared view of significant events, unresolved risk conditions, authorized decisions, and supporting records for security, risk, and executive review.

04

Control & Framework Mapping

Designed to connect control objectives, configured responses, and available evidence to applicable requirements. References do not establish compliance.

Measures and dashboard capabilities are in development; no validated exposure reduction or performance baseline is represented.

Documentation

iS NOT

Control

02 / Core Engines

Three Engines.
One Governance Loop.

Three coordinated capabilities are being designed to connect the basis for an AI action, a proportionate response, and legitimate decision authority.

Explains

Provenance Engine

Designed to preserve relevant context, constraints, rationale, and event history for reconstruction and accountable review.

Tamper-evident AI decision records arranged for provenance and audit review.

Enforces

Intervention Engine

Designed to slow, constrain, escalate, pause, or block proposed actions when configured controls require a response.

Human hand pausing an automated workflow at a secure proportional intervention control panel.

Governs

Custody Engine

Designed to make decision authority enforceable and auditable by defining who sets constraints, approves overrides, and accepts residual risk.

Secure dual-authorization checkpoint representing governed authority over AI actions.

03 / Shared Evaluation Layer

One Action. Multiple Lenses.

Constraint Polyphony is Piscys’ proposed approach to evaluating evidence, proportionality, and decision authority together, surfacing conflicts across Provenance, Intervention, and Custody.

Conceptual diagram showing evidence, risk, and authority constraints connected to three vertically arranged engines—Provenance, Intervention, and Custody—within a shared evaluation scope.
Conceptual diagram showing evidence, risk, and authority constraints connected to three vertically arranged engines—Provenance, Intervention, and Custody—within a shared evaluation scope.

These evaluations are intended to inform proportionate intervention and accountable authority review, with supporting evidence surfaced through AI Trust Metrics and the Assurance Dashboard.

04 / Healthcare Concept Prototype

Explore the Governance Experience

Explore an illustrative healthcare workflow connecting assessment, intervention, accountable authority, and evidence review.

Concept interface. Illustrative data and metrics. Platform in development.

Explore each view. Select the image to enlarge.

Piscys Assurance Dashboard showing governed clinical AI systems, interventions, decision timing, confidence status, and active overrides.
Assurance — See significant events and unresolved conditions.

Concept interface. Illustrative data and metrics. Platform in development.

Explore each view. Select the image to enlarge.

Piscys Intervention Engine evaluating clinical AI confidence drift through four-lens analysis, proportional intervention levels, and counterfactual response options.
Evaluate — Examine constraints and possible responses.

Concept interface. Illustrative data and metrics. Platform in development.

Explore each view. Select the image to enlarge.

Piscys Custody Engine constraint library showing enforceable policy-as-code, version history, dual-key authorization, and runtime enforcement scope.
Authority — Review scoped decision rights.

Concept interface. Illustrative data and metrics. Platform in development.

Explore each view. Select the image to enlarge.

Piscys Custody Engine decision seal showing cryptographic proof, accountable signatory authority, override conditions, and immutable decision evidence.
Seal — Inspect the proposed decision record.

Concept interface. Illustrative data and metrics. Platform in development.

Explore each view. Select the image to enlarge.

Piscys audit trail showing append-only records of AI actions, human authorizations, constraints, cryptographic hashes, interventions, and outcomes.
Audit — Examine the supporting event history.

05 / Initial Industries

Governance Where Decisions Carry Consequences

Financial risk committee reviewing a governed credit-risk decision in an institutional setting.

Financial Services

Examine fraud review and credit-decision exceptions, focusing on required evidence, exception authority, escalation paths, and other consequential AI-assisted workflows.

Clinician reviewing diagnostic imaging in a governed AI-assisted clinical workflow.

Healthcare

Examine prior-authorization review, clinical recommendation routing, and sensitive-data access, focusing on accountable review, evidence requirements, and escalation responsibilities.

Illustrative workflow focus, not customer deployments. Governance design does not establish medical efficacy or regulatory compliance.

06 / Engage Piscys

Start With One Consequential Workflow

A focused executive conversation can clarify the decision, its accountable owner, and the control or evidence questions worth examining.

Our founders bring experience in Microsoft platform security, enterprise program management at Microsoft and Avanade, and site reliability engineering at Cisco.

Meet the founders →

01

Executive Briefing

Frame a priority workflow, identify its principal governance questions, and outline useful next-step options.

02

Rapid Risk Assessment

Develop prioritized findings on control gaps, decision authority, and supporting-evidence requirements within an agreed scope.

03

Implementation Design Sprint

Develop proposed control, escalation, authority, and evidence specifications, including integration and validation requirements.

Advisory engagements are intended to clarify priority workflows, control requirements, and evidence needs, informing platform design and future validation.

Advisory services are available now. Software platform capabilities remain in development.

Help Shape The Platform

We are seeking financial-services and healthcare design partners to help define priority workflows, control requirements, and future platform validation.

For Chief AI Officers, Chief Risk Officers, CISOs, CTOs, and General Counsel responsible for consequential AI workflows.

07 / Operating Approach

Governance Across the AI Lifecycle

Piscys’ approach is informed by the NIST AI Risk Management Framework functions: Govern, Map, Measure, and Manage. Governance informs the lifecycle as systems and operating conditions change.

Govern

Define accountable roles, decision rights, risk tolerances, and escalation policy.

Map

Identify the workflow, its consequential actions, and affected stakeholders.

Measure

Examine relevant risk signals, safeguards, and supporting evidence.

Manage

Prioritize proportionate responses and monitor residual risk as operating conditions change.

Framework references inform design; they do not establish certification or compliance.

Read the NIST AI RMF →

08 / Ecosystem

Supporting Piscys Development

Piscys is a member of the NVIDIA Inception Program, a startup program supporting companies developing innovative technologies.

About the NVIDIA Inception Program →

09 / Contact Piscys

Discuss Your AI Governance Priorities

Bring one consequential AI workflow, its decision owner, and the control or evidence question you want to examine. 

For design partnerships or other inquiries, tell us what you have in mind.

Advisory services available now. Software platform in development.

 

 

 

 

 

 

 

Privacy Policy →

Which workflow are you examining, and what decision, control, or evidence gap matters most?