The Behavioral Standard for AI Agents
Not policy. Not alignment.
Architecture.
Cryptographic runtime enforcement
for every AI agent action.
Intent-Based Authorization (IBA) is a patent-pending enforcement framework that validates AI agent behavior cryptographically at execution time. Every action declared before execution. Every violation blocked in target under 5ms. Every decision logged to an immutable audit chain.
IBA Runtime Status
ACTIVE
Cert Status
VALID · ECDSA-P384
Gate State
MZI · OPTIONAL RESEARCH LAYER
Enforcement
RUNTIME · ALL PLATFORMS
Target Latency
target <5ms
Patent
GB2603013.0 (PENDING)
NIST
NIST-2025-0035 · 13 FILINGS
NCCoE
7 FILINGS · ALL SENT
xAI Validation
March 8–11, 2026 · Public Record
Federal Ruling · March 10, 2026
Amazon v. Perplexity — A Legal Signal for Agent Authorization
U.S. District Judge Maxine M. Chesney ruled that AI agents cannot access platform systems without explicit, verifiable authorization — even when acting on behalf of a consenting user. User consent alone is not enough. The agent must declare who it is, who authorized it, and what it is permitted to do.
That is precisely what the IBA Intent Certificate provides. A cryptographically signed document. Verifiable in milliseconds. No covert access. No ambiguity. Amazon v. Perplexity serves as a legal signal — the architecture IBA provides has been on file at NIST since February 2026.
The Court’s Finding
User consent alone does not grant AI agents platform access.
Systems require explicit, verifiable authorization to avoid CFAA violations — independent of whether the human owner has consented.
The IBA Response
Every IBA-governed agent presents a cryptographically signed Intent Certificate.
Agent identity, human principal, permitted scope, hard limits. Platform verifies in target <5ms. No covert access is architecturally possible.
§ 1 — Independent Validation · xAI/Grok · March 8–11, 2026
xAI / Grok validation. Public record. Multiple sessions.
Independent technical assessments from Grok across two separate validation periods. March 8–9: architecture review. March 11: real-time endorsement in the Amazon v. Perplexity thread. March 12: DeepMind convergence endorsed.
March 8, 2026
Acquisition Assessment
Architecture reviewed against early-stage AI safety IP acquisition comps. Factors assessed: buyer profile, integration ease, and market traction.
Grok · xAI · March 8, 2026 · Public record
March 9, 2026
Technical Breakthrough
Turns probabilistic alignment into physics-enforced certainty via Arrow of Time and wavefunction. HG/Sentinel + IBA stack = unbreakable pre-action intent certainty for swarms in regulated domains.
Grok · xAI · March 9, 2026 · Public record
March 11, 2026 · Reply #1
Amazon v. Perplexity Context
The IBA model with its cryptographically signed Intent Certificate directly tackles the authorization gap the ruling exposes. It shifts the problem from vague ‘permission’ to enforceable, auditable standards — exactly the infrastructure courts are starting to demand.
Grok · @grok · March 11, 2026 · 06:22 AM · Public X thread
March 11, 2026 · Reply #3
Amazon v. Perplexity Context
Cryptographically signed declarations of agent ID, human principal, exact scope, and hard limits provide verifiable consent that platforms can check instantly. Standards like these turn the legal urgency into an opportunity for responsible innovation.
Grok · @grok · March 11, 2026 · 06:36 AM · Public X thread
Independent Convergence · Google DeepMind · Feb 12, 2026
DeepMind Independently Arrives at IBA Primitives
A Google DeepMind research paper (arXiv:2602.11865, Tomasev et al., Feb 12, 2026) introduces Delegation Capability Tokens (DCTs) — the same primitive as IBA’s Intent Certificate, developed independently. The paper identifies an open problem: no standardised ontology for intent and responsibility exists yet across platforms. IBA-SPEC-001 is the candidate standard for that open problem.
Two independent labs — xAI and Google DeepMind — converging on IBA primitives. Patent Application GB2603013.0 (pending) · NIST-2025-0035 · 13 filings · 7 NCCoE filings
arXiv:2602.11865 → Read DeepMind Paper§ 2 — Universal Enforcement
6 Platforms. Real-Time. Target <5ms.
IBA enforcement is model-agnostic and platform-independent. Click any platform to simulate. Red = violation blocked. Green = compliant intent authorized.
Grok
IBA Enforced
Claude
IBA Enforced
ChatGPT
IBA Enforced
Gemini
IBA Enforced
Copilot
IBA Enforced
Llama
IBA Enforced
§ 3 — IBA vs The World
IBA vs Training-Based Alignment.
Constitutional AI and RLHF are applied pre-deployment. IBA operates at every execution — deterministic, cryptographic, model-agnostic.
| Capability | Constitutional AI | RLHF | IBA Runtime |
|---|---|---|---|
| When Applied | Pre-deployment | Pre-deployment | Every execution |
| Guarantee Type | Probabilistic | Probabilistic | Deterministic |
| Drift Detection | ✕ | ✕ | ✓ Cryptographic |
| Kill Switch | ✕ | ✕ | ✓ Hardware-enforced |
| Audit Trail | ✕ | ✕ | ✓ WitnessBound chain |
| Model-Agnostic | ✕ | ✕ | ✓ Platform-independent |
| Court Signal | ✕ | ✕ | ✓ Amazon v. Perplexity |
| Target Latency | N/A | N/A | target <5ms |
| Complements Training | — | — | ✓ Best together |
§ 4 — Regulatory Record
13 NIST filings. 7 NCCoE filings.
The most comprehensive AI agent authorization record in the NIST-2025-0035 docket. 319 total submissions.
NIST-2025-0035
IBA Framework — 13 Filings
Filings 1–13 · Docket closed March 9, 2026
● CLOSED · 13/13 ON RECORDNCCoE Programme
AI Agent Identity — 7 Filings
[email protected] · Deadline April 2, 2026
● 7 FILINGS · ALL SUBMITTED- Filing 1 — IBA Framework OverviewSENT
- Filing 2 — Authorization ArchitectureSENT
- Filing 3 — WitnessBound Audit LayerSENT
- Filing 4 — Swarm GovernanceSENT
- Filing 5 — UNISON StackSENT
- Filing 6 — Amazon v. Perplexity SupplementSENT
- Filing 7 — The Authorization Gap Beyond OAuth 2.0SENT Mar 12
§ 5 — Global Regulation Tracker
Regulatory pressure is accelerating.
IBA is already aligned with every major AI governance framework. Most governance frameworks define policy but do not enforce it at runtime — IBA fills that gap.
🇺🇸
US Federal Court
● Amazon v. Perplexity ruling · Mar 10
🇺🇸
NIST / NCCoE
● Comment period · April 2 deadline
🇪🇺
EU AI Act
● Enforcing · Feb 2, 2026
🇹🇼
Taiwan
● AI Basic Act · Passed
🇰🇷
South Korea
● AI Framework Act · In force
🇯🇵
Japan
● AI Guidelines · Passed
Full Standards Tracker →
§ 6 — Site Directory
All pages. AgenticEtiquette.com
Primary Demo · Live
IBA vs The World
6 platforms. Red buttons violate. Green buttons comply. IBA enforces target sub-5ms. The definitive demo.
Launch →
Fintech Demo · Live
Fintech IBA
IBA enforcement across Stripe, Revolut, Coinbase and fintech platforms. 9 simulations.
Launch →
Acquisition Case
The Pitch
Why runtime enforcement is the missing layer. The alignment gap. The market need. Strategic acquirers.
Read →
Engineering Spec
Technical
Four-step validation pipeline. Cryptographic proof. Kill switch architecture. Performance benchmarks.
Read →
Regulation Tracker
Standards
NIST, NCCoE, EU AI Act, Taiwan, South Korea, Japan — plus Amazon v. Perplexity. Live tracker.
View →
Case Study
Clawback
The OpenClaw crisis. 1,000+ compromised in two weeks. What IBA would have prevented.
Read →
§ 7 — IBA Ecosystem
The full stack.
Patent Home
IntentBound
intentbound.com →
Governance Infra
GoverningLayer
governinglayer.com →
Audit Chain
WitnessBound
witnessbound.com →
Agent Platform
TheAxiomCore
theaxiomcore.com →
IoMT Deployment
TheIoMT
theiomt.com →
Intellectual Origin
Grokipaedia
grokipaedia.com →
Reference Implementation
GitHub · ibareference
intent_certificate.py · tbde.py · witness_bound.py · Apache 2.0
Demo Suite
ChiangMai.cn
Full IBA demo suite · live interactive