Platform Architecture

The Four Layers of AI Governance.

From low-overhead telemetry ingestion to immutable cryptographic proof, Blocklog provides a multi-layer defense against AI audit risks.

LAYER 01 · The Cryptographic Ledger

Blocklog — Immutable Core

Active · Core Layer

Key Question Answered: "Can we prove why this decision happened?"

Append-only governance record store with SHA-256 hash chaining, Merkle tree aggregations, and automated evidence exports.

Append-only record store with hash chaining

Forensic replay & decision reconstruction

Automated DPDP & RBI regulatory export

Ed25519 cryptographic attestations

LAYER 02 · Execution & Orchestration

Active · Ingestion Layer

Key Question Answered: "What exactly did the agent do?"

Deep orchestration telemetry capturing prompt variations, tool call chains, browser actions, and retrieval context in real time.

OpenTelemetry & LangChain native hooks

Stale input & retrieval staleness tracking

Latency & token cost breakdown

Multi-agent handoff tracing

LAYER 03 · Runtime Enforcement

Active · Governance Layer

Key Question Answered: "Should the agent be allowed to do this action?"

Human-in-the-loop gate sitting in the execution path. Holds high-risk financial or data operations until authorized.

Configurable risk threshold triggers

Slack & Webhook approval workflows

Cryptographically signed human overrides

Automated rollback on rejection

LAYER 04 · Confidential Execution

Enterprise VPC Option

Key Question Answered: "How do we safely execute high-risk actions?"

Hardware-enforced confidential computing environment for processing sensitive financial transactions or untrusted LLM outputs.

Hardware enclaves (AWS Nitro / GCP Confidential)

Memory encryption during processing

Zero plaintext exposure to cloud provider

Formal mathematical proof generation

Full documentation for Python, Node.js, REST APIs, and OpenTelemetry pipelines.