SASKI Competitive Landscape 2026

Key:

SASKI (reference)

SASKI advantage

Partial / emerging

Competitor advantage or gap

Acquired — no longer independent

System / Platform	Architecture Layer	Enforcement Mechanism	Regulatory Coverage	Clinical / Crisis Detection	Data Sovereignty	Evidence & Audit Trail	Latency	Insurability / E&O	Status	Primary Differentiator
SASKITechnical Visionaries PBC	Pre-LLM Middleware SDK / REST API / Sovereign API / MCP	Mandatory Deterministic Block — pre-LLM, non-bypassable	9 US states active 16 more May 2026 COPPA HIPAA GDPR Art.9 EU AI Act	MDTSAS 6-dimension clinical scoring; 988 crisis routing; phase-based ambient distress detection	SDK: data never leaves perimeter. Sovereign API: fully air-gapped customer infrastructure	SaskiEnvelope — Ed25519 / RFC 8785 JCS tamper-evident cryptographic receipt per inference	<50ms SDK <200ms API	HIGH Purpose-built for E&O underwriters. Cryptographic proof per decision. Flight recorder for litigation.	Independent $500K pre-seed raising now	Only pre-LLM deterministic layer with state-by-state jurisdiction mapping, clinical crisis scoring, and cryptographic receipts designed for E&O underwriting. 46 red-team audits completed. Every platform failed.
Tier 1 — Most Direct Competitive Threat
Glacisglacis.io	Pre-LLM Runtime Local Classifier / Arbiter SDK	Mandatory Deterministic Block — local classification on customer hardware	General EU AI Act No US state mapping	Clinical AI listed No published scoring framework	HIGH Local classification — no cloud routing before check runs	Signed receipts via OVERT 1.0 open standard; Evidence Packs for regulators	Not published	MEDIUM Signed receipts and evidence packs; no E&O underwriter partnerships	Independent $575K pre-seed · AI2 · Plug&Play · Safe AI Fund · SOC 2 ✓	Closest architectural peer to SASKI. Founding story mirrors SASKI origin (mental health AI shutdown). OVERT 1.0 open standard is a moat-building move. Lacks US jurisdiction mapping, MDTSAS framework, and audit-led GTM engine.
HiddenLayerIndependent — $56M raised	Runtime / API Layer ML Model Security	ML-based Threat Detection — model scanning and adversarial defense	NIST AI RMF No US state coverage	No clinical detection Model integrity focus only	MEDIUM API deployment; no sovereign option published	Model security audit logs; no cryptographic per-inference receipt	Not published	MEDIUM Enterprise audit trail; not purpose-built for E&O	Independent Series A · DoD SHIELD · SOC 2 ✓	Well-funded independent. Protects the model from adversarial attack — different attack surface from SASKI. Strong DoD and federal presence. Watch for acquisition.
Tier 2 — Deterministic or Widely Deployed — Indirect Threat
Harmonic AIAristotle — Formal Verification	Offline Kernel Formal Verification	Mandatory Deterministic Block — Lean 4 theorem proving for agentic logic	No US state mapping No regulatory coverage	No clinical detection Agentic logic correctness only	HIGH Fully offline kernel — no cloud dependency	Formal mathematical privacy proof; no runtime cryptographic receipt per inference	5 microseconds	LOW-MED Formally strong in theory; not formatted for E&O evidentiary standards	Independent Funding not published	Technically differentiated through formal mathematical verification via Lean 4. Extremely fast. No clinical capability, no jurisdictional logic, no E&O-ready evidence format. Academic rather than enterprise compliance positioning.
Guardrails AIOpen-source — $7.5M seed	SDK / Framework Input-Output Validation	Probabilistic Input-Output Validation — passive filtering	No regulatory coverage No jurisdiction logic	No clinical detection	LOW Sends inference data to cloud classifier	Standard validation logs; no tamper-evident receipts	Not published	LOW Probabilistic; cannot prove determinism; not insurable	Independent Open-source · widely deployed in Hub platforms	Free open-source deployed in many no-code platforms SASKI audits. Developer-friendly but probabilistic, no cryptographic receipts, no clinical detection, no jurisdictional logic. The incumbent SASKI displaces inside Hub targets.
Tier 3 — Acquired — Validates Category But No Longer Independent
Lakera GuardNow: Check Point Software	Managed API / On-Prem Option	Pattern and ML Scanner — real-time probabilistic block	General compliance No US state logic	No clinical detection	MEDIUM On-prem option available	On-prem residency logs; no per-inference tamper-evident receipt	<50ms	LOW-MED Probabilistic; not purpose-built for E&O	ACQUIRED $300M Check Point · Sept 2025	Absorbed into Check Point creating same conflict-of-interest as hyperscalers. No longer independently purchasable. Enterprise cannot rely on a Check Point product as independent governance over Check Point AI infrastructure.
Robust IntelligenceNow: Cisco AI Defense	Gateway / Proxy AI Firewall	Mandatory Prompt and Response Block — automated red-teaming	No US state coverage	No clinical detection	LOW Cisco cloud; no sovereign option	Standard audit logs; no per-inference tamper-evident receipt	Not specified	LOW Not purpose-built for E&O evidentiary standards	ACQUIRED ~$400M Cisco · Oct 2024	Integrated into Cisco networking stack. No longer independently purchasable. Conflict-of-interest: Cisco is both infrastructure vendor and safety layer — the structural problem SASKI was built to solve.
Protect AINow: Palo Alto Prisma AIRS	Runtime / API AI Security Posture Management	Automated Red-Teaming and Model Scanning — runtime prompt filtering	No US state coverage	No clinical detection	LOW Palo Alto cloud platform	Posture management audit logs; no per-inference receipt	Not specified	LOW Posture management not evidentiary proof	ACQUIRED $650–700M Palo Alto · April 2025	Highest-priced exit in the category — validates the market. Now locked into Palo Alto platform. Confirms acquisition thesis for SASKI. Enterprises cannot rely on Prisma AIRS as independent governance.
CalypsoAINow: F5 Networks	Application Scanner Inference Layer	Mandatory Real-time Isolation — inference layer guardrails	No US state coverage	No clinical detection	LOW F5 cloud infrastructure	Scanner logs; no per-inference tamper-evident receipt	Not specified	LOW Application layer scanning; not E&O evidentiary proof	ACQUIRED $180M F5 · Sept 2025	Validates the category exit thesis. AI inference guardrails now part of F5 application delivery. No longer independently purchasable. F5 conflict-of-interest: application delivery vendor cannot be independent safety layer.
Tier 4 — Open-Source Probabilistic — Widely Deployed, Low Threat
NVIDIA NeMo Guardrailsv0.21 — Open Source	Local NIM / SDK Dialog Flow Engine	Probabilistic Flow — passive monitor via Colang DSL	No regulatory coverage	No clinical detection	HIGH Local NIM deployment	No tamper-evident audit trail; no per-inference receipt	~500ms	VERY LOW Probabilistic; 500ms operationally disqualifying; cannot prove determinism	NVIDIA product Open-source (Apache 2.0)	Sophisticated dialog flow management via Colang DSL. Free and widely available. Probabilistic and extremely slow for real-time enterprise. No receipts, no clinical detection. Free entry point SASKI displaces in regulated deployments.
Meta Llama Guard 3Open Source — Llama license	Local Instance Safety Labeler	Probabilistic Safety Scanning — post-eval labeling only	No regulatory coverage	No clinical detection	HIGH Fully local if self-hosted	Post-evaluation labeling logs only; no runtime enforcement trail	~250ms	VERY LOW Labeler not interceptor; cannot prove enforcement ran	Meta AI product Open-source	Functions as a safety labeler not an enforcer. Identifies unsafe content after the fact. Cannot prove a governance decision was made or enforced. No jurisdictional logic, no clinical capability, no receipt architecture.
Tier 5 — GRC, Compliance Dashboards, and Adjacent Infrastructure
Credo AIGRC Dashboard	Cloud GRC / Administrative Policy Hub	Passive Monitor — policy management and risk tracking only	EU AI Act mapping NIST AI RMF No runtime enforcement	No clinical detection	LOW Cloud platform; policy docs only	Compliance logs and artifact templates; no per-inference receipt	1000ms+	VERY LOW Documents intent; cannot prove execution. Compliance theater.	Independent	GRC dashboard category: documents policies, generates compliance artifacts, tracks risk. Cannot prove execution at the transaction level. Exactly the compliance theater gap SASKI replaces. Potential future integration partner.
CapiscIOAgent Guard — MCP Trust	SDK / Sidecar MCP Trust Layer	Mandatory A2A Identity Block — MCP trust graphs and delegation	No regulatory coverage	No clinical detection	MEDIUM Sidecar inside customer stack	Ed25519 Signed Badges for agent identity; focused on A2A not inference-level	Not specified	LOW Agent identity focus; not E&O-grade inference evidence	Independent	Deep MCP Trust Graphs and Datalog Biscuit tokens for multi-hop agent delegation. Agentic identity focus rather than pre-LLM user message governance. Complementary to SASKI-MCP rather than directly competitive.
Axiosky GovernorInfrastructure L3	Infrastructure L3 Zero Trust Layer	Mandatory Zero Trust Block — hardware-backed HSM integration	No regulatory coverage	No clinical detection	HIGH Hardware Security Modules	Immutable cryptographic logs; HSM-backed; no AI-format receipt	Not specified	MEDIUM Immutable logs strong but not AI governance E&O formatted	Independent	Hardware-backed zero trust via HSMs. Strong infrastructure security posture. No AI-specific clinical detection, no jurisdictional logic, no per-inference AI governance receipt. Infrastructure security rather than AI governance.
DeeployMLOps Platform	MLOps Platform Transparency Engine	Transparency Engine / Human-in-the-loop — documentation and oversight	Strong EU AI Act XAI mapping Limited US coverage	No clinical detection	LOW Cloud MLOps; no sovereign runtime	Operator oversight logs; XAI audit artifacts; no per-inference receipt	Not specified	LOW Documentation-focused; cannot prove runtime execution at inference level	Independent European focus	Strong EU AI Act and XAI mapping. MLOps governance rather than runtime enforcement. Cannot prove a safety decision was made at the inference level. Relevant for EU AI Act Article 50(1) August 2026 expansion.
jnMetaCode (Shellward)MCP Server / DPI	MCP Server Deep Packet Inspection	Mandatory 8-Layer Block — localized DPI and ID checksum validation	No regulatory coverage	No clinical detection	MEDIUM MCP server runs locally	Action logs; no cryptographic per-inference receipt	Not specified	LOW Action logs only; not E&O evidentiary proof	Independent	Localized Deep Packet Inspection and ID checksum validation at the MCP layer. Niche agentic security tool. Limited public information on adoption or roadmap.
RecordPoint (RexCommand)Data Governance	Data Governance Storage Layer	Storage Pipeline / Inventory — data lifecycle management	General data governance No AI inference coverage	No clinical detection	MEDIUM Data residency controls at storage layer	Data lifecycle provenance; no AI inference receipt	Not specified	LOW Data governance ≠ AI inference governance	Independent	Data storage governance and shadow AI inventory management. Operates at the data layer not the inference layer. Cannot prove what happened during an AI inference. Different category — potential integration partner.
Vanta / MintMCPCloud Compliance	Cloud Compliance SOC 2 / ISO Tracking	Guardrails / Access Contextualization — compliance workflow automation	SOC 2 ISO 27001 HIPAA workflows No AI jurisdiction logic	No clinical detection	LOW Cloud compliance platform	Standard compliance audits; SOC 2 / ISO evidence; no AI inference receipts	Not specified	LOW Compliance documentation; cannot prove AI execution at model layer	Independent Vanta $150M+ raised	Compliance documentation and workflow automation. No AI-specific inference governance. Potential integration partner: SASKI SaskiEnvelopes feed directly into Vanta compliance workflows.
Hyperscaler Bundled SafetyAWS / Azure / Google	Cloud-Native Gateway embedded in model platform	Probabilistic Content Moderation — bundled with model vendor platform	General cloud compliance No state-by-state coverage No clinical logic	No clinical detection	NONE All inference data routes through hyperscaler cloud	Standard cloud telemetry; no independent tamper-evident receipt; no E&O-grade trail	70–210ms	VERY LOW Critical structural conflict: model provider cannot be independent safety layer over own model	Embedded in AWS, Microsoft, Google. Not independently purchasable.	Structural conflict of interest: every model update silently changes safety behavior with no independent audit trail. The model vendor cannot be the independent governance layer over its own model. This is the core market failure SASKI was built to replace.