Technical Specifications & Acquisition Overview — WBA
THE GOVERNANCESTACK.
IBA + WBA. Runtime enforcement plus authority lineage.
The complete patent family for AI governance in regulated industries.
Technical depth for CTOs. Acquisition clarity for corporate development.
Technical Specifications
WHAT WBA DOES
Five architectural layers. Each one patentable. Together they form the first complete authority lineage system for AI governance — designed to survive a courtroom.
ABE — Authority Baseline Engine
Behavioural Principal Profiling
Continuously profiles each signing principal across multiple dimensions. Establishes what normal looks like for every person authorized to declare AI agent intent.
CAD — Contextual Anomaly Detector
Real-Time Signing Event Evaluation
At every signing event, computes a live Anomaly Score comparing current conditions against the principal’s established baseline. Flags deviations before they enter the chain.
WTE — Witness Threshold Engine
Configurable Co-Signing Requirements
Determines how many independent witnesses must co-sign a high-anomaly re-authorization before it enters the chain. Threshold tables are cryptographically locked by governance policy.
SCR — Signing Context Record
Immutable Context Chain
Every signing event — regardless of anomaly level — produces an immutable Signing Context Record. Not just what was signed. The conditions under which it was signed. Cryptographically chained to all prior records.
ERA — Evidentiary Reconstruction API
Regulatory-Grade Lineage Export
At any future date, produces a complete, cryptographically verifiable authority lineage for any agent deployment. Formatted for regulatory submission, legal proceedings, and board review.
IBA + WBA — Combined Stack
The Complete Governance Layer
IBA handles runtime. WBA handles authority lineage. Together they answer every question a regulator, board, or legal team will ask about an AI agent deployment — at any point in the future.
Performance Characteristics
BUILT FOR PRODUCTION
| Metric | Value | Notes | Comparison |
|---|---|---|---|
| Signing event evaluation | <5ms | CAD + WTE combined latency at P99 | Negligible overhead |
| Anomaly score computation | <2ms | Against 90-day baseline profile | Real-time capable |
| SCR creation + chain append | <3ms | SHA-256 + cryptographic chain link | Sub-transaction overhead |
| Full lineage reconstruction | <500ms | Complete deployment history, any depth | Regulatory submission ready |
| Concurrent deployments | Unlimited | Independent chain per deployment | Horizontally scalable |
| Model compatibility | Any | Model-agnostic wrapper — no fine-tuning | GPT, Claude, Gemini, Grok, open-source |
| Runtime overhead on IBA | ~0% | WBA operates on signing layer only, not action layer | No agent performance impact |
Integration
HOW IT FITS
WBA wraps the intent declaration layer of IBA. No changes to the agent itself. No model modifications. No pipeline disruption. Drop-in governance for any deployment.
# Standard IBA intent declaration intent = SignedIntent( principal="[email protected]", scope="patient-record-read", resources=["records/*"], expires="2026-02-20T18:00:00Z" ) # WBA wraps the signing event — transparent to the agent wba_result = WitnessedSigning( event=intent, principal_profile=ABE.get_profile(intent.principal), context=OperationalContext.capture() ) # High-anomaly path — requires co-signing if wba_result.anomaly_score > 0.65: wba_result = WTE.require_witnesses(wba_result, threshold=2) # Signing blocked until witnesses obtained # SCR locked into chain — immutable from this point scr = SCR.lock(wba_result) # Normal IBA runtime proceeds — WBA transparent to agent agent = IBAAgent(intent=wba_result.verified_intent)
# Six months later — regulator asks for authority lineage lineage = ERA.reconstruct( deployment_id="hospital-ai-ward-7", from_date="2026-02-01", to_date="2026-08-20", output_format="regulatory-pdf" ) # Returns: complete chain of every signing event # Who signed. When. Anomaly score. Witnesses obtained. # Cryptographically verified. Tamper-evident. Court-ready. lineage.verify() # → True — chain integrity confirmed lineage.export() # → PDF formatted for regulatory submission
Security Architecture
HOW IBA + WBA ADDRESSES KEY AGENT THREATS
Mapped to NIST RFI threat categories. Every threat has a structural response — not a policy response.
| Threat | NIST RFI Category | IBA + WBA Mitigation | Layer |
|---|---|---|---|
| Coerced / pressure-driven re-signing | Specification gaming · Insider backdoor | CAD flags incident pressure + time-of-day anomalies. WTE requires multi-party co-signing or mandatory delay before SCR is finalised. Coercion leaves a detectable signature. | WBA |
| Silent scope mutation / drift | Misaligned objectives · Coordination drift | Immutable SCR chain captures every authorization state. ERA API reconstructs the full history. Scope drift is detectable at any future audit point — not just in real time. | IBA + WBA |
| Insider authority abuse | Backdoor · Privileged access misuse | ABE establishes baseline behavioural profile at deployment. CAD scores anomaly delta against baseline — high deltas trigger WTE escalation. Insider abuse deviates from ABE baseline. | WBA |
| Indirect prompt injection | Indirect prompt injection | IBA Authorization() validates every action against signed intent. Injected instructions causing trajectory deviation return Authorization = FALSE. Structurally impossible — not filtered. | IBA |
| Data poisoning / rug pull | Data poisoning · Training manipulation | WBA hashes and versions the data state at signing time. Any post-signing mutation is immediately detectable by hash comparison against SCR-bound state. Poisoning cannot accumulate silently. | WBA |
| Supply-chain key compromise | Backdoor · Supply-chain attack | Bitcoin blockchain anchoring provides independent verification. A fraudulent SCR created post-key-compromise has a blockchain timestamp inconsistent with the legitimate authority chain. ERA identifies the exact breach point. | WBA |
| Confused-deputy attack | Privilege escalation · Cross-context abuse | Intent declarations are context-specific — authority granted in one context cannot be exercised in another. Different intent trajectories, different Authorization() evaluations. No ambient inherited authority. | IBA |
Current frameworks stop at cryptographic validity. IBA + WBA adds mandate legitimacy — the gap no existing standard closes.
Architecture Flow
A SIGNING EVENT — STEP BY STEP
Step 01 — Trigger
Intent Declaration Initiated
A human principal initiates a signing event — either creating a new intent declaration or re-signing to mutate the scope of an existing agent deployment. WBA intercepts before IBA processes.
Step 02 — ABE
Baseline Profile Retrieved
The Authority Baseline Engine retrieves the principal’s 90-day behavioural profile. Six dimensions: typical signing times, operational contexts, scope delta history, frequency patterns, co-signatory norms, resource class patterns.
Step 03 — CAD
Anomaly Score Computed
Current signing conditions compared against baseline. Time-of-day deviation, scope delta magnitude, active incident flags, operational pressure indicators all evaluated. Anomaly Score computed in <2ms.
Step 04 — WTE
Witness Threshold Determined
Based on Anomaly Score and resource class classification, the Witness Threshold Engine determines whether the signing can proceed with single authority, requires co-signatories, or requires temporal delay. High-anomaly events are blocked until threshold met.
Step 05 — SCR
Signing Context Record Locked
Regardless of anomaly level — whether the signing proceeded or was blocked — an immutable Signing Context Record is created. Contains all 9 mandatory fields. Cryptographically chained to all prior SCRs for this deployment. Immutable from this point.
Step 06 — IBA
Runtime Enforcement Proceeds
If the signing was approved, the verified intent passes to IBA for runtime enforcement. The agent operates within the declared scope. WBA is transparent to the agent — zero runtime overhead. The authority lineage is permanently recorded.
Acquisition Overview
Pre-grant pricing for IBA + WBA patent family
ACQUIRE NOW · PRE-GRANT · POST-GRANT THE SAME IP COMMANDS A MULTIPLE
WHY NOW — THE WINDOW IS MEASURED IN MONTHS
EU AI Act enforcement begins 2026. Every AI agent in healthcare, finance, legal, and critical infrastructure needs demonstrable authorization before deployment. HIPAA. MiFID II. SEC Reg BI. Taiwan AI Basic Act. The regulation is being enforced now. The patent was filed first.
Acquire IBA + WBA at pre-grant pricing today. Post-grant, the same IP commands a multiple. The window closes when the patent is examined. IBA examination is already in progress from the 5th February filing.
Google / DeepMind — 2014
AI safety + research IP. Pre-revenue, pre-product. Pure IP and team value.
Microsoft / Nuance — 2021
Clinical AI + regulated vertical IP. Healthcare compliance as core asset.
Meta / Manus — 2025
Agentic AI platform acquisition. Agent infrastructure as strategic moat.
IBA + WBA — Asking 2026
Pre-grant. First-mover. NIST filed. xAI validated. Patent family. The authorization layer every regulated AI deployment needs.
For Corporate Development
What You Acquire
Two patent-pending inventions covering the complete AI authorization stack. The only filed framework addressing both runtime enforcement and authority lineage. NIST federal docket position. xAI validation on public record. 10-page live demo platform. GitHub codebase. Two category-defining domains.
For CTOs
What You Deploy
Model-agnostic drop-in governance layer. No fine-tuning required. Works with GPT, Claude, Gemini, Grok, any open-source model. <5ms overhead per signing event. Zero agent runtime impact. Regulatory-grade evidentiary output. Compatible with EU AI Act, HIPAA, MiFID II, SEC Reg BI from day one.
For Legal Teams
What You Prove
Complete cryptographically verifiable audit trail for every AI agent action and every authorization event. Reconstructible authority lineage from any point in the deployment history. Formatted for regulatory submission. Designed for courtroom survivability. The answer to every question a regulator will ask.
For Investors
Why The Timing Is Right
The regulation mandating this technology is being enforced now. The patent was filed before the enforcement deadline. No competitor has filed equivalent IP. The NIST docket position establishes federal standards engagement. xAI independent validation is public record. First-mover advantage is intact — but not for long.
Intellectual Property
THE PATENT FAMILY
Foundation Patent
IBA
GB2603013.0 — Filed 5th February 2026 — UK IPO
Intent-Bound Authorization. Cryptographic binding of AI agent actions to signed human intent. Runtime enforcement, temporal decay, trajectory evaluation. The foundation layer — did the agent act within declared scope?
Companion Patent
WBA
WitnessBound — UK IPO (Primary) — PCT by 5 Feb 2027
Witness-Bound Authority. Authority lineage verification. Contextual anomaly detection, witness thresholds, Signing Context Records, evidentiary reconstruction. The meta-layer — was the scope declaration itself legitimate?
Federal Standards
NIST
NIST-2025-0035 — Tracking: mls-ubpf-pryy — Filed 19 Feb 2026
First formally filed authorization framework in the NIST CAISI AI agent standards process. Federal docket position. xAI independently validated February 2026. Public record.
Federal Standards
ALIGNMENT WITH NIST AI AGENT STANDARDS INITIATIVE
Launched February 17, 2026 — IntentBound Positioned at the Foundation
RFI — Due March 9, 2026
NIST-2025-0035
NIST seeks input on agent threats — specification gaming, misaligned objectives, backdoors, indirect prompt injection. WBA’s CAD anomaly detection + WTE co-signing directly mitigates coercion-induced and pressure-driven authority mutations. Filed 19 Feb 2026. Tracking: mls-ubpf-pryy.
NCCoE Paper — Due April 2, 2026
NCCoE identity and authorization paper probes agent identity proofs and auditing. ERA’s reconstructible, tamper-evident Signing Context Record chains fit the requirement directly — every agent action auditable to a signed, immutable human authority lineage.
NIST Initiative — Feb 17, 2026
Secure · Interoperable · Trusted
NIST’s Initiative focuses on secure agent function on behalf of users, interoperable ecosystems, and public trust. WBA’s blockchain-anchored authority lineage provides provable, non-mutable human authority — the technical substrate NIST’s public trust requirement demands.
IntentBound is already in the federal process — before the window closes.
We welcome collaboration for NCCoE demos or guideline contributions. Contact: [email protected]
The Inventor
I filed the authorization layer they’ll need before their AI doctors are allowed to practice.”
Jeff Williams. 79 years old. Recovering from a heart attack. Living in Chiang Mai, Thailand.
While the labs raced to build agents, he filed the patent that governs them.
ONE CONVERSATION.CHANGES EVERYTHING.
Acquisition enquiries, licensing, and technical collaboration welcome from any jurisdiction.