b
BLOCKLOG

Logs Are Claims.
Blocklog Makes Them Proof.

Block financial AI actions unless policy-approved and cryptographically authorized. Replay decisions, verify integrity, and prove what autonomous systems actually did.

Join 30-Day PilotSee Proof Flow

Hash Chain Integrity

99.999%

daily verification window

Proof Verification

30 Sec

median retrieval time

Built for financial systems, audits, and autonomous agents.
SOC2AI GovernanceAudit TrailsProof SystemsRuntime Authorization

Forensic report preview

production-derived

Decision ID

dec_9af1c18

refund-agent-v2 · 2026-05-25T09:42:11.882Z

Policy State

policy.v17.3

thresholds · staleness gates · human checkpoints

Input staleness tablemilliseconds at decision time
risk_score22018 msstale
fraud_signal_count0 msfresh
customer_tier391 msfresh
prior_disputes5220 mswarning

Approval lineage

KYC checkpoint complete · Manual fraud review skipped · Senior approval not required

Counterfactual summary

Decision would have routed to human review if risk_score freshness were below 5s.

EU AI Act Article 12 ✓ · SR 11-7 ✓ · CFPB ✓ · signed hash: 7a8e1b... · chain position: 2281184

// The Incident

$180,000. 72 Hours. 400 Decisions. Zero Governance Record.

Meridian Payments ran an AI refund adjudication agent against a customer risk score that was 22 seconds stale. Three fraud signals never reached the model. By the time the investigation started, the causal chain had already dissolved into application logs, screenshots, and partial institutional memory.

Blocklog would have captured the stale inputs, the exact policy version, the approval path, the execution token, and the counterfactual trigger that would have changed the outcome.
Hour 00
Customer score fetched
22 seconds stale
Hour 08
Fraud signal delayed
Not delivered to model
Hour 21
Refund adjudication
Checkpoint skipped
Hour 48
Execution completed
$180,000 exposed
Hour 72
Investigation starts
No governance record
EU AI Act · August 2, 2026 · 75 days
Colorado AI Act · June 30, 2026
SR 11-7 · Federal Reserve · Active
DORA · EU · Active

// Proof Flow

Every financial AI action becomes independently verifiable.

Runtime Authorization

Block refunds, payouts, and high-risk AI actions unless policy-approved and cryptographically authorized.

Policy EngineApproval GatesToken VerificationExecution Control
agent.request()→
policy.match()→
token.sign()→
processor.execute()✓

Forensic Replay

Reconstruct exactly why an AI system made a financial decision, including policy state, stale inputs, approvals, and causal lineage.

Timeline ReplayCausal GraphInput StalenessDecision Lineage
input.score: stale→
policy v17.3→
checkpoint skipped→
counterfactual delta✓

Cryptographic Proof

Hash chains, Merkle proofs, signed evidence bundles, and independently verifiable audit exports.

Hash ChainingMerkle RootsSigned BundlesIndependent Verification
2cb7f3...→
7c9a8d...→
67ab1e...→
root: 4de0a1...✓

The Report Your Regulator Will Actually Accept.

Generated from production data in under 20 minutes. No engineer required in the room.

Enterprise PDF export

Forensic Compliance Report

Decision ID and timestamp

Agent identity fingerprint

Policy version evaluated

Approval lineage completed or skipped

Counterfactual summary

Regulatory mapping

Decision ID · dec_9af1c18
Agent fingerprint · agt_f84e1a2
Policy version · refund-policy-v17.3
Input staleness · 4 fields analyzed
Approval lineage · 3/4 checkpoints completed
Counterfactual · stale risk_score flipped review path
Regulatory mapping · EU AI Act Article 12 ✓ · SR 11-7 ✓ · CFPB ✓
Cryptographic attestation footer · signed hash 7a8e1b... · chain position 2281184

// Entry Flow

Shadow Mode to Enterprise Contract in 30 Days.

Step 1
Engineer installs SDK

npm install @blocklog/sdk. Shadow mode activates with zero production impact.

Step 2
Blocklog observes every AI financial decision

Captures inputs, timestamps, policy version, approval chain status, and escalation triggers.

Step 3
Forensic report generated after 30 days

Real dollar amounts. Real compliance implications. Real decisions from production.

Step 4
Compliance officer reads the report

The output is readable by compliance and risk without engineering archaeology.

Step 5
Enterprise contract signed

Authorization gate enabled. Execution dependency established. Infrastructure pricing locked.

Your AI is Making Decisions. Can You Prove What It Did?

Install shadow mode in under 5 minutes. No production impact. No procurement approval. First forensic report in 30 days.

Join 30-Day PilotRead the Docs