The Problem Deploy Demo OWASP FAQ
EU AI Act Article 12 · August 2026 OCC Bulletin 2025-26 · In force now NY RAISE Act · Jan 2027 HIPAA § 164.312(b) · OCR enforcement active

Your AI is already
making decisions.
Prove every one
of them.

A Decision Passport is a cryptographically signed, tamper-proof record of every consequential AI agent action: who authorized it, what policy applied, and exactly why. Signed by an independent third party. Not your cloud. Not your own infrastructure. Hand it to the OCC examiner, the SOC2 auditor, the EU supervisory authority. Nothing in your current stack produces this. When your examiner asks "what controlled that decision?" This is what you show them.

Two deployment modes: capture every LLM call with zero code changes, or prove every regulated decision with one decorator. Both produce the same artifact. See deployment modes →

The gap

The regulator asks why.
You have nothing.

Banks, insurers, and healthcare companies are deploying AI agents to approve loans, flag fraud, deny prior authorizations, triage patients, and more. When the regulator asks why the agent did what it did, most companies cannot answer.

OCC examination scenario

Your AI agent approved $4.2M in refunds last quarter. The OCC examiner wants three things: which policy version governed each decision, whether the model received exactly what was authorized, and proof the record was not touched afterward. You have logs. They are mutable. They are not enough.

Litigation scenario

Outside counsel on the other side subpoenas your AI decision records. Your audit logs are not write-protected (WORM). Your database rows can be updated. Your audit trail has no cryptographic signature. Everything you hand them can be challenged on authenticity. Without cryptographic proof of integrity, you cannot demonstrate the record was not modified after the fact. And if your cloud provider signed the evidence about decisions that ran on their own infrastructure, opposing counsel will challenge the independence of that record too.

HIPAA Audit Scenario

Your AI agent flagged 12,000 patients for care gap outreach last quarter. HHS OCR opens an investigation. They want the audit trail: which agent decision identified each patient, what data it accessed, under which policy version, with what authorization. Your logs show the agent ran. They do not show what it was authorized to access, or whether the input it received was exactly what was approved. HIPAA § 164.312(b) requires audit controls that can prove what your AI accessed and whether it was authorized. With mutable logs, you cannot prove that. Your stack produces none.

Current toolWhat it doesWhat it lacksAgentOS fills
CloudTrail / audit logsRecords API callsNo decision causality, no cryptographic seal. Even immutable cloud logs fail the independence test. The same provider cannot be both the infrastructure and the auditor.Single cryptographically signed artifact per AI agent decision, signed by an independent third party
AI monitoring platformsTraces execution, monitors driftMutable logs, not legal evidenceWORM-committed, RFC 3161 timestamped
OPA / policy enginesEvaluates rules at runtimeNo artifact, no causality chainPre-execution binding + cryptographically signed artifact
LLM observability toolsLLM observability and loggingMutable database logs, no cryptographic proofEd25519 signed, independently verifiable
SIEM / log aggregatorsAggregates logs across systemsLogs are mutable, not sealedEd25519 signed, WORM-committed, designed to withstand adversarial examination

LLM observability tools

Built for debugging and operational visibility, not legal proceedings. Mutable logs. No cryptographic signature. No WORM commitment. Outside counsel cannot establish integrity in discovery.

AI monitoring platforms

Built for operational awareness, not legal proceedings. An OCC examiner will not accept a mutable data export as audit-grade proof, regardless of format.

The artifact

The Decision Passport.

One self-contained artifact per AI agent action. Signed, sealed, timestamped.
Click any field below to see what it proves. ↓

Each field exists for a forensic reason. Click to see it.

What TIER-A means HSM-signed · WORM-committed · RFC 3161 timestamped. TIER-A is the evidence grade designed for named compliance controls. HSM-signed, WORM-committed, RFC 3161 timestamped.
decision_passport_dp_9f3a2c1e.json TIER-A

  "passport_id"    "dp_9f3a2c1e-4b7d-4e8a-b1c2"
  "schema_version" "1.1.0"
  "tier"           "TIER_A"HSM · WORM · RFC3161
  "capture_layer"  "BUSINESS_LAYER"decorator
  "consequential"  true

  "principal"
    "agent_id"  "refund-approval-agent-v2"
    "tenant_id" "tenant_acme_fintech"
  

  "action"
    "type"  "REFUND_APPROVAL"
    "decision_ts" "2026-04-12T09:23:41.847Z"
  

  "policy_evaluation"
    "result"       "ALLOW"
    "policy_id"    "refund-policy-v4.2.1"
    "input_hash"   "sha3_256:a3f82c1d..."pre-exec sealed
  

  "causality_chain"
    "step_sequence" 3
    "total_steps"   3
    "prev_step_hash" "sha3_256:9e8d7c6b..."tamper-evident
  

  "signing_attestation"
    "algorithm"  "Ed25519"
    "hsm_level"  "FIPS_140-2_LEVEL_3"production HSM
    "signature"  "3045022100a1b2c3..."
  

  "storage"
    "worm_commit"   "CONFIRMED"S3 COMPLIANCE
    "rfc3161_ts"    "APPLIED"after WORM
    "customer_bucket" trueyou own this
    "lock_verification_latency_ms" 187
  

  "compliance_controls"
    "OCC_BULLETIN_2025-26" "NYDFS_PART_500"
    "EU_AI_ACT_ARTICLE_12" "SOC2_CC6.1"
    "NIST_SP800-53_AU-10"
  

  "verification_uri" "https://verify.getagentos.ai/dp_9f3a2c1e"

Verify independently: agentos verify dp_9f3a2c1e

Integration

Two ways to deploy.

Two deployment modes. Both produce the same Decision Passport. Start with auto-instrument for immediate coverage. Promote regulated paths to the decorator.

Mode 1  ·  Decorator
@agentos.consequential
01

Wrap the regulated function

One decorator on the function that makes the regulated decision. No changes to internal logic or return types.

BUSINESS_LAYER  ·  consequential=true
02

Pre-execution seal

SHA3-256(JCS(business_inputs)) committed into a signed Authorization Token before the function runs. Tampered input returns 403, zero artifact.

OCC  ·  NYDFS  ·  Litigation-ready
03

Decision Passport issued

Ed25519 signed. WORM-committed. RFC 3161 timestamped. Verifiable independently using the standalone CLI and your S3 bucket.

TIER-A  ·  Synchronous
Mode 2  ·  Auto-instrument
agentos run
01

Zero code changes

Run agentos run python app.py or set AGENTOS_AUTO_INSTRUMENT=1. Same pattern as ddtrace-run.

LLM_LAYER  ·  consequential=false
02

Client boundary capture

Every LLM API call intercepted at the client library boundary. SHA3-256(JCS(LLM_API_request)) committed as input_hash.

Article 12  ·  GDPR  ·  NIST AI RMF
03

Decision Passport per call

Ed25519 signed. WORM-committed. RFC 3161 timestamped. Complete audit trail. Same standalone verify CLI as the decorator.

TIER-A  ·  Zero code change
Compliance coverage

Nine regulatory frameworks.
One artifact.

Each passport names the specific controls it addresses. Not general positioning. Named fields, named controls, named regulations.

In force now · Exam priority
OCC Bulletin 2025-26
Model risk management for AI-assisted decisions at national banks

OCC Bulletin 2025-26 directs national banks to govern AI agent decisions as model risk. Examination cycles now include AI governance readiness.

The Decision Passport closes the evidence gap: policy version, input hash, principal identity, all signed and WORM-committed.

Addresses: model_version_hash · policy_version_hash · signing_attestationRead bulletin ↗
Active enforcement risk
NYDFS Part 500
AI systems for NY-licensed covered entities

NY-licensed fintechs and insurers face NYDFS enforcement risk for AI agent decisions without audit trails.

The Decision Passport provides the cryptographic evidence NYDFS examiners can verify independently.

Addresses: policy_version_hash · signing_attestation · WORM storageRead regulation ↗
August 2, 2026 deadline
EU AI Act Article 12
Automatic logging for high-risk AI systems · US fintechs with EU customers are caught

Article 12 requires tamper-proof automatic logging for high-risk AI, with post-hoc reconstruction capability.

The Decision Passport captures every LLM call via auto-instrument mode with zero code changes.
Deadline is firm.

Addresses: causality_chain_hash · full_prompt_snapshot · capture_layerRead Article 12 ↗
In force now
GDPR Article 22
Accountability for automated decisions affecting individuals

Article 22 governs automated decision-making affecting individuals and requires documented accountability.

The Decision Passport provides that documentation: which model, which policy, which input, which output. All signed and independently verifiable.

Addresses: principal_identity · policy_version_hash · execution_resultRead Article 22 ↗
Effective Jan 1, 2027 · NYDFS enforcement
NY RAISE Act
Signed December 2025 · Consequential AI decision accountability

The NY RAISE Act requires frontier AI developers to publish safety frameworks and report safety incidents. Enterprises deploying those models need documentation of which model version governed each decision.

The Decision Passport captures model_version_hash and registry_validation_status for every decision. The documentation enterprises need to show they used compliant model versions.

Addresses: consequential · capture_layer · BUSINESS_LAYER artifactsRead Act ↗
SOC2 + Federal standard
SOC2 CC6.1/CC4.1 + NIST SP 800-53
Non-human identity authorization · AU-10 Non-repudiation · COSAiS Agentic Overlay

CC6.1 maps to authorization_token_id. CC4.1 maps to execution_result. NIST AU-10 requires non-repudiation of principal actions.

The Decision Passport uses Ed25519 digital signatures to provide the non-repudiation mechanism this control requires. AgentOS is designed to align with these emerging controls.
The COSAiS project (NIST) is developing SP 800-53 control overlays for autonomous agent systems.

Addresses: AU-10 · authorization_token_id · Ed25519 + WORM + RFC 3161Read SP 800-53 ↗
Controls AgentOS addresses
SOC2 CC6.1 / CC4.1
Authorization of non-human identities · Processing integrity

SOC2 CC6.1 requires documented authorization controls for non-human identities. CC4.1 requires evidence of processing integrity.

The Decision Passport proves non-human identity authorization via signed authorization_token_id for CC6.1. CC4.1: execution_result and policy_evaluation prove processing integrity.
Both controls are documented for auditor review.

Addresses: authorization_token_id · execution_result · policy_evaluationRead trust criteria ↗
In force · Rule update expected 2026
HIPAA § 164.312(b)
Audit controls for AI agent access to electronic protected health information

The proposed 2025 Security Rule amendments would significantly strengthen audit logging requirements for AI agent interactions with ePHI. Under the proposed amendments, every agent decision touching patient data would require a record of who authorized it, what input it received, and what it produced. Mutable logs do not satisfy this requirement.

The Decision Passport captures the hash of inputs, not the inputs themselves, keeping PHI out of the artifact.

Addresses: principal_identity · authorization_token_id · execution_resultRead Security Rule ↗
In force · FDA regulated industries
21 CFR Part 11
FDA electronic records and audit trails for AI in drug development, clinical trials, and medical devices

Any AI system that creates or modifies GxP records must log which model version acted, on what input, producing what output, with a timestamp.

The Decision Passport captures all of this in a single signed artifact, designed to address Part 11 audit trail requirements for AI-driven decisions in pharmaceutical and clinical contexts.

Addresses: model_version_hash · input_hash · execution_result · rfc3161_tsRead CFR Part 11 ↗
Security review

OWASP Agentic AI Top 10 mapping.

All 10 risks mapped to specific Decision Passport fields. Hand this to your security team. The review checklist can close in one meeting rather than 2-3 weeks. The controls live in code.

#OWASP RiskEvidence TypeAgentOS Controls
1Goal hijackingPrev + Recordinput_hash + substrate enforcement + AUTHORIZATION_INPUT_MISMATCH
2Tool misusePrev + Recordtool_call_log + authorization_token_id + action_type
3Identity abuseRecordingprincipal_identity + signing_attestation
4Memory poisoningRecordingfull_prompt_snapshot + rag_evidence_bundle + causality_chain_hash
5Cascading failuresRecordingcausality_chain_hash + instrumentation_coverage + PASSPORT_COMMIT_FAILED
6Insecure output handlingRecordingoutput_pre_postprocess + execution_result
7Code execution abusePrev + Recordaction_type + tool_call_log + BLOCK_ON_SERVICE_UNAVAILABLE
8Human-agent trust exploitationPrev + Recordprincipal_identity + policy_version_hash + availability_policy
9Supply chain risksRecordingmodel_version_hash + registry_validation_status + sdk_binary_sha3
10Rogue agentsRecordingdeployment_mode + evidence_tier + instrumentation_coverage
Ed25519FIPS 186-5 Signing
FIPS 140-2Level 3 HSM
RFC 3161Trusted Timestamp
29 FieldsSchema v1.1.0
34Design Issues Closed
10/10OWASP Agentic AI Mapped
About the founder
SS
Founder, AgentOS
Swapan Shridhar

19 years systems engineering  ·  HP (OpenVMS, Tru64 UNIX)  ·  VMware ESX  ·  FedRAMP GovCloud  ·  Apache Ambari PMC

Started at HP. Filesystem work on Tru64 UNIX, then backup and restore on OpenVMS using tape libraries. Tape is write-once by design. Once written, it stays. S3 Object Lock in AgentOS enforces the same property, different medium.

Moved to VMware and spent years in the ESX storage stack implementing T10 DIF/DIX. That is the standard that proves data was not silently corrupted between host and storage device. Same problem, one layer down.

Later: FedRAMP GovCloud delivery, Control Plane and Workload side. FIPS 140-3 cryptographic compliance. Evidence packages for federal authorising officials, produced firsthand.

19 years, same question: how do you prove the data is exactly what it should be, and that nobody changed it? AgentOS asks that question about AI decisions.

💾
HP: Tru64 UNIX filesystem & OpenVMS backupFilesystem development and backup/restore using tape libraries. Write-once storage at the OS level. Same property AgentOS enforces in S3 Object Lock.
🔐
VMware ESX: T10 DIF/DIXData Integrity Field and Extension in the ESX storage stack. End-to-end proof data was not corrupted between host and device.
🏛️
FedRAMP GovCloud deliveryControl Plane and Workload side. NIST 800-53 controls, authorising official interactions, evidence production. Built it, did not consult on it.
🔒
FIPS 140-3 compliant images and processCryptographic module validation at the level US federal agencies require. Not self-certification.
🔗
Chainguard supply chain hardeningHardened base images, supply chain security in production. Maps to OWASP Agentic AI Risk 9.
📐
Apache Ambari PMCProject Management Committee member. Public, peer-reviewed open source infrastructure work.
Common questions

Frequently asked.

For regulated decision functions, the right starting point is the decorator (@agentos.consequential) on the functions that make regulated decisions. This produces the strongest evidence, designed to meet OCC examiner, NYDFS, and litigation discovery standards.

Auto-instrument is the right starting point if your immediate deadline is EU AI Act Article 12 or GDPR Article 22. It captures every LLM call without code changes.

We recommend running both in production. A ContextVar coexistence guard ensures exactly one passport per call, never two.
Yes. Same Decision Passport schema (v1.1.0), same Ed25519 signing, same S3 Object Lock COMPLIANCE mode WORM, same RFC 3161 timestamp, same standalone agentos verify CLI. The capture_layer field tells the auditor which forensic claim applies. The verify CLI accepts both v1.0.0 and v1.1.0 artifacts.
SDK integration in a clean development environment: 30 minutes. Enterprise production integration: 1-4 hours. Enterprise security review of the ABSA interceptor: 3-6 weeks, reduced by the OWASP mapping. Total onboarding to first production artifact: 4-8 weeks.
In production, TIER-A means HSM-signed inside FIPS 140-2 Level 3 hardware, WORM-committed, and RFC 3161 timestamped. Design partner phase uses software signing with the same Ed25519 algorithm. Designed as the primary evidence grade for SOC2 auditors and OCC examiners.

TIER-B: Customer-hosted attestation with a declared 60-second uncertainty window. Corroborating evidence. The distinction is explicit in the evidence_tier field of every artifact.
Yes. The OWASP Agentic AI Top 10 mapping will be published open source on GitHub (docs/owasp-mapping.md, May '26). Every control claim is tied to a specific named field in the Decision Passport schema, also published. Your security team can review the mapping against the schema and verify every claim independently.

The mapping distinguishes between Prevention and Recording. Naming what each mode actually does is more honest than claiming prevention for everything.
Yes. OCC Bulletin 2025-26 is in force now and is an active examination priority. Examination cycles now include AI governance readiness. The bulletin extends model risk management requirements to cover AI agents making consequential decisions at national banks.

Whether the OCC cares is settled. The real question is whether your model risk program documents what your AI agents are doing, under which policy version, with what inputs, and whether those records can survive examination. Most cannot today.
Your artifacts live in your S3 bucket under your Object Lock policy. AgentOS has no write access post-provisioning. Every artifact stays in your possession, in your bucket, under your retention policy.

The standalone verify CLI will be open source. Your auditor verifies authenticity using only your bucket access and the published public key. No AgentOS cooperation required. We are establishing escrow custody of source code and keys for the insolvency scenario.
The Decision Passport is written to your S3 bucket in your AWS account. The signing operation receives the hash of your data, not the data itself. AgentOS has no write access to your bucket after provisioning. The customer_bucket field in every artifact confirms your ownership.
OpenTelemetry spans are mutable, infrastructure-dependent observability records designed for debugging. They can be deleted, modified, or selectively exported. An OCC examiner or opposing counsel can challenge their integrity.

Production artifacts are signed inside a FIPS 140-2 Level 3 HSM. Design partner phase uses software signing with the same Ed25519 algorithm. Both are written to S3 Object Lock COMPLIANCE mode and timestamped by a neutral RFC 3161 TSA. The artifact is designed to withstand adversarial examination. One is for debugging. The other is for regulators. Not the same thing.
The sealing operation is asynchronous. Your agent function returns before the WORM commit completes. The passport_id is synchronously returned sub-10ms. The WORM commit, RFC 3161 timestamp, and DynamoDB ledger write happen in the background and never block your agent's execution.

The lock_verification_latency_ms field documents the WORM sealing time per artifact. In production, values are typically 150-300ms.
Because you would be signing your own evidence. An OCC examiner or opposing counsel will ask: who holds the signing key? If the answer is the same company that made the decision, or the same cloud that hosted it, the independence argument fails.

AgentOS holds the signing key. The timestamp comes from a neutral third-party TSA. Your S3 bucket holds the artifact. No single party controls all three. That separation is what makes the record credible to a regulator who was not in the room when the decision was made.
Get started

One conversation.

For Compliance, Legal & Risk Teams
Validate the artifact against your audit requirements.

Looking for 1-2 design partners to build the Decision Passport to your auditor's specification. You define what it must prove. We build to your requirements.

90-day pilot · $10-25K · 50% upfront · Design partner pricing guaranteed for the first two customers.

OCC examination readiness, EU AI Act Article 12, or HIPAA audit controls.

Request Access Book a 30-minute call

contact@getagentos.ai

For Engineering Teams
Evaluate the architecture.

The Decision Passport schema is open. The OWASP mapping is public. The standalone verify CLI will be open source.

SDK integration: 30 minutes in a clean development environment. Enterprise security review: 3-6 weeks, reduced by the OWASP mapping.

No account needed to evaluate the schema and verify a sample artifact.

Explore Decision Passport Demo See Deployment Modes

SDK integration: 30 minutes in a clean environment · Enterprise production: 1–4 hours

Request access.

90-day pilot · $10-25K · 50% upfront. One conversation to see if it fits.

Or book directly: 30-min call ↗

Received. We will be in touch within one business day.