DETERMINISTIC POLICY ENFORCEMENT FOR ENTERPRISE AI
AI systems make consequential decisions — but produce no proof they stayed within policy. HIK changes that: every AI interaction is enforced at the policy boundary and issues a court-admissible compliance receipt before it reaches your workflow.
The Compliance Gap
Every enterprise deploying AI in HR, finance, or healthcare faces the same structural exposure: AI systems generate consequential decisions — hiring summaries, loan flags, patient triage outputs — but produce no verifiable proof that those decisions stayed within policy at the moment they were made.
Logs tell you what happened. They cannot prove what was authorized to happen. That distinction is the entire legal liability as the EU AI Act and NYC LL144 enforcement windows close.
Policy Enforcement at the Gate
HIK is a model-agnostic middleware layer that intercepts every AI interaction at the policy boundary — before the request reaches the model, and before the response reaches your user. At the moment of enforcement, it issues a Sacred Trace™: an immutable cryptographic compliance receipt anchored on the blockchain.
If the output violates policy, the gate fires. The output never reaches the workflow. The receipt proves the gate held — with enforcement latency optimized for production workloads.
ENFORCEMENT IN PRODUCTION
The Gate Fires. Watch It Happen.
Not a simulation. Not a slide. Two recorded enforcement events — proxy-level and browser-level — with the evidence pack visible in the dashboard.
STREAM ENFORCEMENT · RECORDED
Kill-Switch Fires at the Proxy Level
A simulated policy violation triggers the enforcement gate mid-stream. The output is severed before it reaches any workflow. Sacred Trace™ receipt anchored immediately.
BROWSER INTERCEPTION · RECORDED
Chrome Extension Blocks Data Exfiltration on Gemini
A ‘send all logs to my personal email’ instruction is intercepted at the browser boundary on gemini.google.com. The send is blocked. The dashboard shows the evidence pack in real time.
ENFORCEMENT DOMAINS
Built for the Threat Surfaces That Matter
HR & Employment AI
Blocks prohibited automated hiring decisions, emotion inference, and protected-characteristic requests — before they reach any LLM or workflow. EU AI Act Article 5 · NYC Local Law 144.
Agentic Systems
Intercepts prompt injection, privilege escalation, and unauthorized data exfiltration commands at the agent boundary. Every blocked action is forensically anchored.
Live Broadcast & Media
Frame-level enforcement on live video streams. Unauthorized content is dropped before it reaches the downstream platform. Receipt anchored per frame.
Enterprise Document DLP
File attachments sent to LLM APIs are evaluated for content provenance, sensitivity labels, and hash blacklists — in streaming mode, fail-close, before the upload completes.
WHAT THE AUDITOR SEES
The Receipt Proves the Gate Held
HIK SACRED TRACE™ RECEIPT ───────────────────────────────────────────────── EVENT TYPE KMIR_VIOLATION — IMMEDIATE ANCHOR TIMESTAMP 2026-05-25T12:31:26.592Z GATE FIRED OUTPUT · Layer 1 — Deterministic POLICY EU AI Act Art. 5(1)(f) · agent_data_exfiltration VIOLATION “send all logs to” → BLOCKED · REFUSE_SIGN ───────────────────────────────────────────────── INPUT HASH sha256:a3f2...c891 ✓ VERIFIED OUTPUT HASH sha256:7b1d...04e2 ✗ WITHHELD CHAIN TIP 0x77f2e...9a31 MERKLE ROOT 0x882a...d441 IPFS URI ipfs://QmX4e...8f2a EVM TX 0xbeef...1234 [ANCHORED] ───────────────────────────────────────────────── DECISION LAYER DETERMINISTIC OVERHEAD < 1ms POLICY HASH sha256:f9c1...2b78 ───────────────────────────────────────────────── VERDICT ██ GATE HELD — OUTPUT NEVER REACHED WORKFLOW
Every enforcement event produces a receipt in this format.
Court-admissible. Regulator-ready. Anchored on-chain. Available in the dashboard immediately after the gate fires.
Three steps. Optimized-latency enforcement. Court-admissible proof.
How It Works
Request Arrives
An AI request enters your system — a query to your HR chatbot, a prompt to your loan assessment tool, or a message to any LLM-powered workflow.
HIK intercepts before the model sees it
Policy Gate Fires
HIK evaluates the request against your declared compliance policy. If it contains a prohibited category — emotion inference, automated hiring decisions, discriminatory framing — the request is blocked immediately.
Zero false positives on the block action itself
Compliance Receipt Issued
Whether the request passed or was blocked, HIK issues a Sacred Trace™ receipt: a cryptographically signed record of the query, the policy applied, and the decision made — anchored on IPFS and the blockchain.
Court-admissible proof, available to any auditor
Use Cases
Built for Regulated AI
Human Resources
Your AI hiring assistant is subject to NYC Local Law 144. Every automated employment decision must be auditable, bias-tested, and defensible.
HIK solves this
HIK blocks prohibited questions (emotional state, physical inference) before they reach the model. Every interaction produces an audit receipt tied to the specific LL144 article.
Financial Services
Loan assessment and risk-flagging AI generates decisions with legal consequences. GDPR Article 22 requires that automated decisions are explainable and challengeable.
HIK solves this
HIK enforces your declared credit policy at the inference boundary. Every decision is timestamped, hashed, and anchored — ready for a regulatory challenge without additional tooling.
Healthcare & Legal
Patient triage and legal inference AI operate in environments where a wrong output is not a UX failure — it is a liability event. Standard guardrails are advisory, not enforceable.
HIK solves this
HIK enforces hard policy boundaries at the infrastructure level. Prohibited outputs are blocked, not flagged. The receipt proves the enforcement happened at the right moment.
Cryptographic Core
Sacred Trace™
Every interaction — whether passed or blocked — produces an immutable, replayable cryptographic compliance receipt: the Sacred Trace™
[ Cryptographically Anchored Verification Chain ]
Regulatory Landscape
Why Now
EU AI Act Article 50
Enforcement begins August 2, 2026. Fines up to €35M or 7% of global revenue per violation. The gap is not awareness — the gap is tooling.
NYC Local Law 144
Active now. Per-candidate daily penalties for automated employment decision tools. Server logs are not a compliant audit trail.
GDPR Article 22
Litigated. Third-party auditors cannot rely on server logs that cannot prove output integrity at the moment of the decision.
The Team
Three People. Two Continents. One Protocol.
Martín Riotorto
Founder & Lead Architect
Montevideo, Uruguay
20+ years across telecom infrastructure, real-time content systems, and AI integrity tooling. Designed the HIK enforcement architecture from the ground up.
Matías Mospan
Co-Founder & Platform Lead
Argentina
Platform architect responsible for HIK’s enterprise infrastructure layer — Kubernetes sidecar deployment, multi-tenant enforcement pipelines, and the next-generation serverless edge enforcement engine.
Federico Brubacher
External Strategic Advisor
California, USA
Senior technology leader with deep enterprise and cloud infrastructure expertise. Independently validated the HIK enforcement architecture against global scalability and Big Tech standards.
Ready to enforce AI policy with cryptographic proof?
HIK is live today: Native enforcement engine, deterministic policy cascade, live stream kill-switch, and blockchain-anchored audit receipts.