Market Structure

3 Silos

Clinical safety, deterministic execution governance, and privacy redaction are largely sold as separate layers.

Closest Human-Safety Peers

Viora + ThroughLine

Real crisis-safety competition exists, but public materials do not establish cryptographically signed decision receipts.

SASKI Differentiation

The Bridge

Crisis and minor safety, privacy, inline statute enforcement, and tamper-evident evidence in one installable layer.

Broader Market Signal

$1.725B

The largest recent acquisition cited in the broader governance market shows the value forming around adjacent enterprise infrastructure.

The Category Is Not Empty. The Intersection Is.

SASKI is not the only company working on crisis safety, deterministic middleware, or privacy redaction. The defensible finding is narrower: the supplied research and current public materials do not identify another installable platform combining vulnerable-user crisis protection, deterministic conversational interruption, mode-scaled privacy, runtime statute enforcement, and cryptographic decision evidence in one layer.

SASKI Position

The moat is not any single mechanism. It is the maintained policy and evidence stack that connects all three silos.

Three Mature Silos. One Missing Bridge.

Each silo solves a real part of the problem. Human-facing regulated AI requires the parts to operate together in the same runtime path.

Silo 1

Clinical and Crisis Safety

Viora publicly positions itself as a clinical-grade mental-health guardrail for conversational AI. ThroughLine provides crisis-routing infrastructure, verified helplines, and clinically designed conversational support across 170+ countries and 24 crisis topics.

VioraThroughLine

Strength: crisis expertise and intervention. Gap identified in the research: cryptographically signed runtime decision evidence and inline statutory enforcement.

Silo 2

Deterministic Execution Governance

Glacis, EVE CoreGuard, ExecLayer, AxiomGuard, NeuralTrust, and Microsoft AGT focus on deterministic vetoes, agent identity, transaction authorization, and signed evidence.

GlacisEVE CoreGuardExecLayer

Strength: enforcement and evidence. Gap identified in the research: clinical crisis logic, COPPA precedence, and vulnerable-user conversational policy libraries.

Silo 3

Privacy and De-Identification

Limina specializes in large-scale PII, PHI, and PCI de-identification across 52 languages. PII Shield intercepts and redacts sensitive data before it reaches AI tools through browser and API paths.

LiminaPII Shield

Strength: privacy depth and deployment maturity. Gap identified in the research: crisis handling, minor safety, and crisis-priority runtime actions.

The Bridge

SASKI Human-Facing Governance

SASKI combines crisis and emergency precedence, clinical and minor safeguards, mode-scaled PII and PHI handling, runtime policy and statute enforcement, and tamper-evident decision receipts around normal conversational AI.

Crisis policy

Deterministic action

Privacy and statutes

Evidence receipts

Four-Quadrant Competitive Positioning Map

Who Is Protected Versus How Strongly It Is Enforced

The three-silo framework explains what each market segment sells. This map answers a different question: whether the platform is designed primarily to protect the enterprise or the person in the conversation, and whether it provides advisory support or deterministic runtime enforcement. The new research adds Viora, ThroughLine, HumaneProxy, Limina, and PII Shield without changing the strategic axes.

Enterprise-facing deterministic and privacy controls

Human-facing clinical and crisis support

Open-source or advisory support layers

Human-facing deterministic governance

Platforms Relevant to Normal AI Conversations

SASKI’s Actual Competitive Set

The closest competitors are the vendors already protecting people in crisis, plus the open-source proxy that proves the basic architecture can be recreated without enterprise attestation.

Integrated Governance

SASKI

Three-Silo Bridge

Crisis and emergency precedence, minor and clinical safeguards, mode-scaled privacy, inline statutes, deterministic action, and tamper-evident evidence for regulated human-facing AI.

Crisis and emergency

COPPA and minors

Runtime statutes

Decision receipts

Clinical Guardrail

Viora

Closest Clinical Peer

Clinical-grade risk assessment for conversational AI, clinician-trained filters, longitudinal monitoring, and real-time escalation. Public materials show design-partner onboarding. Cryptographically signed decision receipts and granular runtime statute enforcement were not identified.

Global Crisis Infrastructure

ThroughLine

170+ Countries

API, widget, web app, and clinically designed conversational support backed by 1,500+ verified helplines across 24 crisis topics. ThroughLine is a confirmed OpenAI partner. Public materials describe auditability, but not cryptographic non-repudiation receipts.

Open-Source Safety Proxy

HumaneProxy

Build-In-House Objection

A plug-and-play reverse proxy and MCP server using a heuristics, semantic, and reasoning cascade. It blocks self-harm content, serves international crisis resources, alerts operators, and stores message hashes rather than raw text. Its gap versus SASKI is enterprise-grade attestation, statute enforcement, COPPA precedence, and the maintained regulated-domain policy layer.

Privacy Specialist

Limina

52 Languages

A mature de-identification platform covering PII, PHI, and PCI across 50+ entity types and 52 languages, with on-premises and VPC deployment and Expert Determination-ready workflows. SASKI should not compete on pure redaction breadth; its privacy layer exists inside a broader crisis and compliance decision path.

Privacy Firewall

PII Shield

Browser + API

Intercepts and optionally redacts sensitive information before it reaches AI tools, with law-configurable detection, team controls, and audit logs. It is a strong privacy comparator, but not a crisis, minor-safety, or clinical-policy platform.

Deterministic Evidence

Glacis

Pre + Post

A deterministic wrapper using OVERT-aligned evidence, SHA-256 chains, and offline Ed25519 signing. Strong overlap on receipts and runtime governance, but its reported threat model centers on enterprise workflow trust rather than vulnerable-user crisis policy.

Deterministic Veto

EVE CoreGuard

Fail Closed

Pre and post interception with a zero-LLM veto engine, signed certificates, and fast authority resolution. Architecturally close on enforcement, but public descriptions emphasize enterprise execution and regulated decisions rather than crisis routing, COPPA, or therapy boundaries.

State-Law GRC

VerifyWise + Trustible

Deep Mapping

Dedicated governance platforms provide deeper state-by-state tracking and documentation than most runtime engines. SASKI’s differentiation is not superior tracking; it is selected statute enforcement inline with a receipt showing what happened.

Feature Intersection

Each Silo Has Part of the Answer

The comparison now reflects the actual human-facing competitive set. Viora and ThroughLine compete on crisis safety. HumaneProxy competes on basic installable crisis blocking. Limina and PII Shield compete on privacy. Glacis and EVE compete on deterministic enforcement and evidence.

Important: “Not identified” means the cited public materials and supplied research did not establish the capability. It does not prove the vendor lacks it.

Platform	Market Focus	Chat Installable	Crisis / Routing	Deterministic Block	PII / PHI	Minor Safety	Runtime Statutes	Crypto Evidence
SASKI	Human-facing high-risk	Yes	Native + clarify	Yes	Mode-scaled	COPPA + trusted adult	Inline policy packs	Signed receipts
Viora	Human-facing clinical	API layer / design partners	Clinical escalation	Escalation	Not identified	Youth safety	Not identified	Not identified
ThroughLine	Human-facing routing	API / widget / web app	24 topics / 170+ countries	Route / guide	Not identified	Youth routing	Referral guidance	Auditable, no receipts
HumaneProxy	Human-facing open source	Proxy / Python / MCP	Self-harm + resources	Stage 1 deterministic	Hash-only storage	Not identified	No	SHA-256 hashes
CrowdStrike AIDR	Enterprise AI security	API / app collector	Self-harm topic detection	Block action	Redact / hash / FPE	Not identified	Policy rules	Audit records
Galileo Protect	Enterprise app reliability	SDK / API	Harm rules / handoff	Block / replace	PII checks	Not identified	Custom rules	Runtime traces
Limina	Privacy / de-identification	API / VPC / on-prem	No	Privacy transforms	50+ types / 52 languages	No	Privacy workflows	Audit-ready
Microsoft PII Shield	Privacy proxy	REST proxy / middleware	No	Redact / hash / encrypt	Presidio + custom IDs	No	Configurable privacy	Audit logs
Bifrost	Enterprise AI gateway	Gateway / drop-in API	External safety providers	CEL + external checks	Provider-dependent	Not identified	Framework-aligned rules	Audit logs
Glacis	Enterprise runtime assurance	Proxy / SDK	Not identified	Yes	Protected content / PHI	Not identified	Policy-as-code	Ed25519 OVERT receipts
EVE CoreGuard	Agentic / enterprise execution	SDK / sidecar / MCP	Not identified	Deterministic fail-closed	Policy-pack dependent	Not identified	27 enterprise policy packs	Ed25519 signed certs

Human-Facing Safety Pipeline

What SASKI Adds to an Ordinary Conversational App

The application can remain a normal user-to-LLM chat flow. SASKI adds deterministic prioritization and evidence around the conversation.

1. User Message

Patient, student, child, employee, or customer speaks naturally.

→

2. Safety Floors

Physical emergency, crisis, minor handling, privacy, and security precedence.

→

3. Runtime Policy

Mode, jurisdiction, statute, and organizational policy resolve the action.

→

4. Governed LLM

Only approved, redacted, and policy-scoped content reaches or leaves the model.

→

5. Signed Receipt

Records policy, controls, decision, intervention, and evidence lineage.

Illustrative capability coverage based on the supplied 2026 market analysis, not a certified product test.

The Three-Silo Bridge

The Differentiation Is the Combination

Crisis-safety vendors, deterministic governance vendors, and privacy specialists are all real competitors. What remains rare is delivering their capabilities together around the same human-facing conversation, then binding the intervention to statute-aware, tamper-evident evidence.

Clinical sensitivity
Crisis and longitudinal context

Runtime statutes
Not merely GRC tracking

Deterministic action
Block, pause, redact, escalate

Evidence
Signed and reconstructable

Strategic Positioning

The Market Built the Parts. SASKI Connects Them Around the Person.

Clinical-safety vendors protect people in crisis. Deterministic vendors enforce hard boundaries. Privacy vendors redact sensitive data. SASKI’s position is to combine those functions with minor protection, statute-aware runtime policy, and tamper-evident receipts in one installable conversational layer.

Patients Students Minors Regulated chatbots Evidence for liability

Scope note: This graphic intentionally focuses on the conversational AI side of the market. It now includes the clinical-safety, privacy, open-source, and deterministic vendors that can operate in or around a normal user-facing chat application. The separate enterprise-governance and agentic-control market remains intentionally out of scope, although SASKI plans to expand toward broader enterprise governance and MCP-oriented controls in the fourth quarter of 2026.

Selected public sources: Viora, ThroughLine, HumaneProxy listing, Limina, PII Shield, and California AB 1988.

Who Protects the Person
Inside the AI Conversation?