SWT3 Clearing Protocol and Shared Responsibility

Name: SWT3 AI Witness SDK
Author: Tenable Nova LLC

Effective Date: August 22, 2026

Purpose

This addendum defines who is responsible for what when you use SWT3 Clearing Levels to manage how your compliance evidence and AI inference data is retained, transmitted, or disposed of.

The SWT3 protocol gives you control over how much data leaves your environment. That control comes with a clear division of responsibility between you and TeNova. This document spells out that division so there is no ambiguity about what is protected, what is kept, and who is accountable.

A. The Four Clearing Levels

When you configure the SWT3 AI Witness SDK or the Axiom platform, you choose a Clearing Level. That choice determines what data is sent to TeNova and what stays on your systems.

Level 0 - Retain

Everything is transmitted and stored: cryptographic hashes, numeric factors, model identifiers, performance data, token counts, and contextual metadata such as provider name, guardrail configuration, and system fingerprint.

Important: no raw prompts or responses are transmitted at any clearing level. That is true at Level 0 and every level above it.

Best for: Development, internal testing, and environments where you need a full forensic trail for later review.

Level 1 - Standard

Cryptographic hashes, numeric factors, model identifier, and contextual metadata are transmitted. This is the default level and the right starting point for most organizations.

Best for: Production environments that need compliance witnessing with a solid audit trail, without exposing unnecessary operational detail.

Level 2 - Sensitive

Cryptographic hashes, numeric factors, and the model identifier are transmitted. All contextual metadata is stripped before anything leaves your environment. The provider name, guardrail names, and system fingerprint stay with you.

Best for: Environments handling regulated data where even metadata could reveal sensitive operational patterns. If your data classification policy restricts what can leave your boundary, this is likely your level.

Level 3 - Sovereign

Only numeric factors are transmitted. The model identifier is hashed before transmission so the original value never leaves your environment. All other identifying information is destroyed before it hits the wire. After the SWT3 Witness Anchor is sealed, the local system purges its temporary memory.

Best for: Air-gapped systems, high-sensitivity environments, and organizations that require zero-knowledge proof of compliance. At this level, TeNova operates as a purely stateless verification provider. We see three numbers and a 12-character fingerprint. Nothing more.

B. Shared Responsibility

What TeNova is responsible for

Protocol integrity. The SWT3 fingerprint algorithm is mathematically sound, deterministic, and one-way. The fingerprint cannot be reversed to reconstruct your data.
Anchor permanence. Once a SWT3 Witness Anchor is sealed and written to the compliance ledger, it cannot be altered, backdated, or deleted by anyone, including TeNova.
Clearing enforcement on our side. When you select a Clearing Level, our platform and SDKs enforce that level at the network boundary. Data excluded by your clearing level is never transmitted to, received by, or stored on TeNova systems.
Enclave integrity. We maintain the cryptographic chain that binds all anchors for a given period into a tamper-evident record. If any anchor is altered after the fact, the chain breaks and the tampering is detectable.
Open verification. The SWT3 protocol specification and the libswt3 reference implementation are published and auditable. You do not need to take our word for any of the above. You can verify it yourself.

What you are responsible for

Choosing your clearing level. You select the level that matches your data classification, regulatory requirements, and risk posture. TeNova does not make this decision for you, and different parts of your organization may choose different levels for different workloads.
Factor custody (Levels 2 and 3). At Clearing Level 2 and above, you hold sole custody of some or all of the data needed to re-verify an anchor. If you lose that data, the anchor becomes mathematically un-verifiable. TeNova cannot recover what we never received.
Factor security. You are responsible for the confidentiality, integrity, and availability of any factors, metadata, or evidence that stays on your systems. If your factors are compromised, exposed, or corrupted, TeNova is not liable for unauthorized re-verification or evidence tampering.
SDK configuration. You are responsible for correctly setting the clearing level in the SWT3 SDK. A misconfigured level may send more or less data than you intend.
Downstream submissions. If you submit SWT3 anchors, compliance artifacts, or evidence packages to any regulatory body, auditor, or reviewing authority, you are responsible for the accuracy and completeness of that submission. TeNova provides the evidence. You own the submission.

C. What TeNova Never Receives

Regardless of which clearing level you select, the following data never leaves your systems:

Raw prompts and responses. The SWT3 protocol works with cryptographic hashes that are computed locally. Your actual AI inputs and outputs are never transmitted to TeNova at any level.
Training data. TeNova does not access, collect, or process your model training data.
Encryption keys. TeNova does not manage, store, or hold your encryption keys.

This is not a policy decision. It is an architectural guarantee.

The SDKs are open source and you can verify this by reading the code.

D. The Self-Custody Trade-off

Clearing Levels 2 and 3 provide maximum privacy by design. We built them that way intentionally because holding your sensitive data creates risk for both of us. But that privacy comes with a trade-off: at higher clearing levels, you become the sole custodian of the evidence needed to prove your compliance history.

We designed this so that both parties are protected. Here is how the responsibility divides at each level.

For Clearing Level 2 (Sensitive)

You acknowledge that TeNova does not retain contextual metadata (provider name, guardrail configuration, system fingerprint) for anchors created at Clearing Level 2. Re-verification of the anchor remains possible using the factors you hold. If you lose those factors, the anchor itself is still intact on our ledger, but you will not be able to independently prove what it represents. This means that a third-party auditor or assessor reviewing Level 2 anchors will need you to provide the retained factors for independent verification.

For Clearing Level 3 (Sovereign)

You acknowledge that for anchors created at Clearing Level 3, TeNova operates solely as a stateless verification provider. You hold sole custody of all factors and metadata required for re-verification. TeNova does not maintain backups of factors or metadata for Sovereign-tier anchors. If your factors are lost, corrupted, or compromised, the corresponding anchor becomes mathematically un-verifiable. TeNova assumes no liability for data loss, evidence gaps, or the inability to provide compliance proof that results from factor loss at this level.

For all Clearing Levels

You assume sole responsibility for the confidentiality, integrity, and availability of all data retained on your systems, including factors, metadata, raw evidence, and encryption keys. TeNova's liability is limited to the integrity of the SWT3 protocol, the correctness of the fingerprint algorithm, and the permanence of anchors stored on TeNova systems, as described in the primary Terms of Service.

Factor Backup Guidance

For organizations operating at Clearing Level 2 or 3, we recommend the following practices for protecting factor custody. These are recommendations, not requirements. Your organization's data protection policies should govern your specific approach.

Maintain multiple copies. We recommend at least three copies of retained factors: one primary and two offsite backups in geographically separate locations.
Encrypt at rest. We recommend AES-256 encryption or equivalent for all stored factor data, with key management separate from the factor store itself.
Sync regularly. We recommend daily backup synchronization for active environments generating new anchors at Levels 2 and 3.
Test restoration. We recommend quarterly restoration tests to verify that backed-up factors can successfully re-verify their corresponding anchors.

E. Incident Response and Breach Notification

If TeNova experiences a security incident that could affect the integrity of anchors or any metadata stored on our systems, we will use reasonable efforts to notify affected clients within 96 hours of confirmed discovery. The notification will include: what data was affected, which clearing levels are impacted, and what actions (if any) you should take.

For clients operating at Clearing Level 3, a breach on TeNova's side exposes only numeric factors (such as latency in milliseconds, token count, and confidence score) and hashed model identifiers with no identifying context. Without the corresponding procedure identifier and tenant context held on your systems, these numbers are not meaningful. We will still notify you, but the practical exposure is minimal by design.

If you experience a security incident and need access to metadata stored on TeNova systems (Levels 0 and 1), we will cooperate with your incident response team and provide relevant records in a timely manner, subject to our Privacy Policy and applicable law.

TeNova does not currently hold SOC 2 Type II certification. We are evaluating options for independent security review of the clearing engine. The SWT3 protocol specification and SDK source code are publicly auditable.

F. Anchor Corrections and Revocation

SWT3 Witness Anchors are permanent records of a compliance fact observed at a specific point in time. They are not editable and they do not expire.

However, mistakes happen. If you discover that an anchor reflects an incorrect verdict (for example, an attestation that should not have been made), the anchor is not deleted. Instead, a revocation record is appended to the ledger referencing the original anchor. The original anchor remains in the cryptographic chain for integrity purposes, but is marked as revoked.

This means the compliance record is always complete and honest. An auditor can see: "This anchor was created on this date, and it was revoked on this date, for this reason." Nothing is hidden or rewritten.

TeNova does not revoke anchors on your behalf. You initiate the revocation, and the platform records it.

Each revocation includes a reason code from the following standard set:

unspecified - general revocation without a specific category
model_recall - the underlying model has been recalled or withdrawn
policy_violation - the inference violated an organizational policy
data_contamination - the input data was contaminated or invalid
consent_withdrawal - a data subject withdrew consent
regulatory_order - a regulatory authority ordered the revocation
error_correction - the original verdict was issued in error

Revocation records are transparent and auditable. An auditor can query the ledger to see all revocations, their timestamps, reason codes, and the original anchors they reference.

G. Anchor Validity Over Time

Anchors are permanent, but compliance is not. An anchor from six months ago tells you what was true six months ago, not what is true today.

It is the responsibility of the reviewing party (whether that is an auditor, an assessor, a regulator, or your own internal team) to decide whether an anchor's age is acceptable for their purposes. TeNova does not impose expiration dates because different industries, frameworks, and jurisdictions have different expectations for how current evidence needs to be.

We recommend running scans at a frequency that matches your compliance cadence. For most organizations, this means daily or weekly collection with periodic review.

Common Retention Expectations by Framework

The following are common retention expectations across regulatory frameworks. Consult your compliance counsel for requirements specific to your organization. This table is provided as general guidance and does not constitute legal advice.

Framework	Typical Retention
NIST 800-53 / FedRAMP	3 years
CMMC v2.0	3 years
EU AI Act	Lifecycle of AI system
HIPAA	6 years
SOC 2	1 year
SR 11-7 / OCC 2011-12	7 years
FAA Part 135 / EASA Part-ML	24 months

Your subscription tier determines platform-side retention (Open: 7 days, Pro: 90 days, Enclave: 365 days, Sovereign: unlimited). If your regulatory framework requires longer retention than your tier provides, maintain local backups or consider a higher tier.

H. Regulatory Applicability

The SWT3 Clearing Protocol is designed to be framework-agnostic. It addresses data retention, disposal, and custody principles that appear across multiple regulatory environments, including but not limited to:

NIST 800-53 and NIST 800-171 (U.S. federal and defense)
CMMC v2.0 (U.S. defense industrial base)
FedRAMP (U.S. cloud services)
HIPAA (U.S. healthcare)
PCI-DSS (payment card industry)
SOC 2 (service organization controls)
EU AI Act (European Union artificial intelligence regulation)
NIST AI RMF (AI risk management)
ISO/IEC 42001 (AI management systems)
SR 11-7 / OCC 2011-12 (U.S. financial model risk management)
FAA Part 135 / EASA Part-ML (aviation AI and autonomous systems)

TeNova publishes separate compliance overlay documents that map clearing levels to specific controls within each framework. See our guides for the overlay relevant to your environment.

The SWT3 protocol specification, fingerprint formula, and SDK source code are publicly available. Organizations subject to EU AI Act conformity assessment or other regulatory review may reference these published materials. For cooperation beyond published documentation, contact legal@tenovaai.com.

I. Related Documents

Factor Handoff Protocol - how factors are securely transferred to your custody at Clearing Level 2 and 3
CMMC Compliance Overlay - clearing level mappings to CMMC and NIST 800-171 controls

J. Contact

Tenable Nova LLC
Email: legal@tenovaai.com
Web: sovereign.tenova.io

This addendum supplements and is incorporated into the TeNova Axiom Terms of Service. In the event of a conflict between this addendum and the primary Terms of Service, this addendum controls with respect to the SWT3 Clearing Protocol. All other terms remain in full force.

Patent Pending. SWT3 and Sovereign Witness Traceability are trademarks of Tenable Nova LLC.