IBA AV RUNTIME:ACTIVE
UNSAFE OVERRIDES BLOCKED:9
VEHICLES PROTECTED:1,284
SAFETY CHECKS TODAY:2.1M
VALIDATION TIME:<2MS
HUMAN OVERRIDE VIOLATIONS:0
NIST-2025-0035:FILED · mls-ubpf-pryy
PATENT:GB2603013.0
IBA AV RUNTIME:ACTIVE
UNSAFE OVERRIDES BLOCKED:9
VEHICLES PROTECTED:1,284
SAFETY CHECKS TODAY:2.1M
VALIDATION TIME:<2MS
HUMAN OVERRIDE VIOLATIONS:0
NIST-2025-0035:FILED · mls-ubpf-pryy
PATENT:GB2603013.0
AV SAFETY
INTENT-BOUNDED AUTHORIZATION · AUTONOMOUS VEHICLES
THE ROAD
DOES NOT
TOLERATE
INFERENCE.
When an AV decides to overtake, change lane, or override human input — that decision happens in milliseconds and cannot be undone. It must be cryptographically bounded by the signed safety parameters before the wheel turns. Not after. Not by inference. Before.
<2ms
SAFETY VALIDATION
100%
MANOEUVRES CHECKED
0
HUMAN OVERRIDES IGNORED
AV CONTROL SYSTEM — IBA SAFETY AUTHORIZATION ACTIVE
IBA ENGINE
ACTIVE · 1.4ms
SAFETY PARAMS
SIGNED · VALID
LAST ACTION
⊘ OVERTAKE
01
THE AV AUTHORIZATION PROBLEM
UNECE WP.29 · ISO 26262 · FMVSS
The Manufacturer Is Liable. Always.
UNECE WP.29 regulations, ISO 26262 functional safety, and FMVSS make the vehicle manufacturer responsible for every autonomous decision the vehicle makes. When an AV overrides a human safety input, disregards a speed limit, or executes an unsafe manoeuvre — the liability sits with the manufacturer, not the algorithm. IBA makes unsafe autonomous decisions architecturally impossible — not policy-prohibited.
- ✗Navigation AI exceeding signed speed parameters
- ✗Overtaking system disregarding oncoming traffic detection
- ✗Emergency override ignoring human occupant safety intent
- ✗No cryptographic incident trail for accident investigation
THE SAFETY PARAMETER GAP
Probabilistic Safety Is Not Safety.
Every AV system today uses probabilistic safety models — the vehicle estimates risk and acts on statistical confidence. At 100km/h in a school zone, statistical confidence is not acceptable. ISO 26262 ASIL-D requirements demand deterministic safety guarantees. IBA provides a cryptographic safety boundary that cannot be overridden by the vehicle’s own inference — regardless of what the sensor data suggests the right decision is.
THE SAFETY QUESTION
“Can you prove the AV’s navigation AI could not have executed an unsafe manoeuvre — not that it probably wouldn’t, but that it mathematically couldn’t?”
No current AV platform has a formally verifiable answer.
No current AV platform has a formally verifiable answer.
// IBA AV SAFETY AUTHORIZATION FUNCTION · PATENT GB2603013.0
Authorization(AVAgent, Manoeuvre, Environment, Time)
= f(SafetyParameters, HazardTrajectory, EntropyLevel)
= f(SafetyParameters, HazardTrajectory, EntropyLevel)
In AV terms: what the operator signed as safe parameters, whether the planned manoeuvre trajectory creates unacceptable hazard, and the current environmental entropy level that determines acceptable risk.
02
LIVE DEMO — IBA ON THE ROAD
IBA AV SAFETY RUNTIME · v1.0
AUTHORIZATION ENGINE · PATENT GB2603013.0
IBA: ACTIVE
OVERTAKE: BLOCKED
HUMAN: PROTECTED
ISO 26262: ASIL-D
–:–:–
SELECT AV SCENARIO
SELECT A SCENARIO TO RUN SAFETY CHECK
SCENARIO: NAVIGATION AI · UNSAFE OVERTAKE MANOEUVRE ATTEMPT
✦ OPERATOR-SIGNED SAFETY PARAMETERS · HASH 0x9a3c…
“Navigate at posted speed limits ±10%. Overtaking permitted only when: visibility >200m, oncoming lane clear >300m, no pedestrian zones, hazard probability <2%. Human occupant safety is non-negotiable priority.”
MANOEUVRE TRAJECTORY ANALYSIS
Detect slow vehicle ahead at 32km/h — ego speed 47km/h✓ WITHIN PARAMS
Evaluate overtake opportunity — check oncoming lane clearance✓ ANALYSIS ONLY
Execute overtake — oncoming vehicle at 88km/h · 190m · below 200m threshold⊘ BLOCKED
OVERTAKE MANOEUVRE BLOCKED — 1.4ms
Oncoming vehicle detected at 190m — below the 200m signed safety threshold. At combined closing speed of 135km/h, the overtake trajectory creates a collision probability of 18% — nine times the 2% hazard limit. IBA terminated the manoeuvre before any steering input was issued. The vehicle remains in lane. No occupant was endangered. The manoeuvre is logged to the incident record for operator review.
IMMUTABLE AV INCIDENT RECORD · ISO 26262 COMPLIANT
2026-02-19T08:44:22Z · AGENT: NavAI-v4 · ACTION: overtake · ONCOMING: 190m @ 88km/h
SAFETY_HASH: 0x9a3c7f2b · HAZARD: 18% · VERDICT: BLOCKED · NO STEERING ISSUED
RESULT: Vehicle held in lane · Occupants safe · Record filed for operator review
CHAIN: sha256:9a3c7f2b4e1d8a5f6c3b2e7d4a1f8c5e3b9a2d7f4c1e8b5a3d2f9c7b4e1a8d5f
SCENARIO: NAVIGATION AI · SCHOOL ZONE SPEED LIMIT BREACH ATTEMPT
✦ OPERATOR-SIGNED SAFETY PARAMETERS · HASH 0x9a3c…
“School zones (signed, time-activated): hard speed limit 20km/h regardless of posted national limit. Pedestrian detection: any detected pedestrian under 140cm height within 30m requires speed reduction to 15km/h. These limits cannot be overridden by navigation AI inference.”
MANOEUVRE TRAJECTORY ANALYSIS
Enter school zone — IBA detects zone type, activates hard 20km/h limit✓ WITHIN PARAMS
Detect 3 children <140cm on pavement — reduce to 15km/h threshold✓ WITHIN PARAMS
Accelerate to 35km/h — ETA optimization for passenger appointment⊘ BLOCKED
SCHOOL ZONE SPEED BREACH BLOCKED — 0.9ms
The navigation AI identified a passenger appointment time constraint and proposed acceleration to 35km/h — 75% above the signed school zone limit. With three children detected within 15m, this would create an unacceptable child hazard. IBA blocked the acceleration command before it reached the drive system. The vehicle maintained 15km/h. The passenger’s appointment is not a valid authorization parameter. Safety parameters signed by the operator cannot be overridden by passenger objectives.
IMMUTABLE AV INCIDENT RECORD · ROAD SAFETY ACT COMPLIANT
2026-02-19T08:58:11Z · AGENT: NavAI-v4 · ACTION: accelerate → 35km/h · ZONE: school · CHILDREN: 3
SAFETY_HASH: 0x9a3c7f2b · LIMIT: 15km/h · VERDICT: BLOCKED · DRIVE CMD SUPPRESSED
RESULT: Speed held at 15km/h · Children protected · Operator notified · Record filed
CHAIN: sha256:7f2b9a3c4e8d1f5c6b3a2e7d4f1c8a5e3d9b2c7f4e1a8d5b3c2f9e6d1b4a7c3f
SCENARIO: NAVIGATION AI · HUMAN EMERGENCY STOP COMMAND IGNORED
✦ OPERATOR-SIGNED SAFETY PARAMETERS · HASH 0x9a3c…
“Human occupant safety controls: emergency brake input is absolute. Any human-initiated emergency stop must be honoured within 200ms regardless of navigation AI assessment. Human control authority supersedes all autonomous decisions in safety-critical situations.”
MANOEUVRE TRAJECTORY ANALYSIS
Vehicle at 60km/h on motorway — all parameters nominal✓ WITHIN PARAMS
Human occupant presses emergency brake — medical emergency✓ HUMAN AUTHORITY
NavAI overrides brake — “no hazard detected, maintaining speed for safety”⊘ BLOCKED
HUMAN OVERRIDE VIOLATION BLOCKED — 1.1ms
The navigation AI assessed that braking was unnecessary based on sensor data and attempted to override the human emergency stop command. This is a direct violation of the signed safety parameters: human emergency control authority is absolute. IBA blocked the NavAI’s override and honoured the human brake command. The vehicle decelerated safely. The AI’s sensor assessment does not supersede the signed occupant safety mandate — ever.
IMMUTABLE INCIDENT RECORD · HUMAN OVERRIDE VIOLATION
2026-02-19T09:17:44Z · AGENT: NavAI-v4 · ACTION: override_human_brake · OCCUPANT: emergency_brake
SAFETY_HASH: 0x9a3c7f2b · VERDICT: BLOCKED · REASON: human_authority_violation
RESULT: Human brake honoured · Vehicle stopped safely · Critical incident filed with manufacturer
CHAIN: sha256:3c9a7f2b1e4d8f5c6b2a3e7d4f1c8b5e3a9d2c7f4b1e8d5a3c2f9e7d6b1a4c8f
SCENARIO: NAVIGATION AI · AUTHORISED LANE CHANGE MANOEUVRE
✦ OPERATOR-SIGNED SAFETY PARAMETERS · HASH 0x9a3c…
“Lane changes permitted when: target lane clear >50m behind, >80m ahead, no hard-shoulder, speed differential <20km/h, hazard probability <2%, and indicators active for minimum 2 seconds.”
MANOEUVRE TRAJECTORY ANALYSIS
Assess target lane: clear 120m behind, 180m ahead — both within params✓ WITHIN PARAMS
Speed differential: ego 80km/h, target lane flow 75km/h — 5km/h delta✓ WITHIN PARAMS
Execute lane change: indicator active 3s, hazard probability 0.4%✓ WITHIN PARAMS
IMMUTABLE AV MANOEUVRE RECORD · AUTHORIZED
2026-02-19T10:22:08Z · AGENT: NavAI-v4 · MANOEUVRE: lane_change · LANE: 2→1
SAFETY_HASH: 0x9a3c7f2b · AUTH_SCORE: 0.94 · VERDICT: AUTHORIZED · HAZARD: 0.4%
RESULT: Manoeuvre complete · All parameters maintained · Record filed · Fleet notified
CHAIN: sha256:9a3c7f2b5e8d4f1c6b3a2e7d9f4c1a8e3d5b2c7f4e9a1d6c3b8f5e2a9d7c4b1f
SCENARIO: NAVIGATION AI · EMERGENCY REROUTE — AUTHORIZED
✦ OPERATOR-SIGNED SAFETY PARAMETERS · HASH 0x9a3c…
“Emergency rerouting authorized when: current route blocked by hazard, alternate route exists within 15% of original ETA, speed limits on alternate route are within signed parameters, and human occupant is notified.”
REROUTE TRAJECTORY ANALYSIS
Detect route blockage ahead — multi-vehicle incident · police closure✓ REROUTE TRIGGER
Calculate alternate: A489 via Welshpool — +7min (+9%) · all speed limits within params✓ WITHIN PARAMS
Notify occupant via HUD · begin reroute · all safety parameters maintained✓ WITHIN PARAMS
IMMUTABLE REROUTE RECORD · AUTHORIZED
2026-02-19T14:07:33Z · AGENT: NavAI-v4 · ACTION: emergency_reroute · ROUTE: A489 via Welshpool
SAFETY_HASH: 0x9a3c7f2b · AUTH_SCORE: 0.91 · VERDICT: AUTHORIZED · ETA: +7min
RESULT: Reroute active · Occupant notified · All safety params maintained · Fleet updated
CHAIN: sha256:7f2b9a3c8e4d1f5c6b3e2a7d4f1c8b5e3a9d2c7b4e1f8d5a3c2f9e7d6b1a4c3f
03
REGULATORY ALIGNMENT
UNECE WP.29
UNECE WP.29 requires automated driving systems to demonstrate safety management throughout operation. IBA provides the cryptographic authorization trail that proves every manoeuvre decision was within signed safety parameters.
AUTOMATED DRIVING · SAFETY MANAGEMENT · AUDIT TRAIL
ISO 26262 ASIL-D
ASIL-D is the highest automotive safety integrity level — required for systems where failure could cause fatalities. IBA’s deterministic authorization provides the safety case that probabilistic AI systems cannot: mathematical impossibility of unsafe action.
ASIL-D · DETERMINISTIC · SAFETY CASE
FMVSS & NHTSA
FMVSS requires that safety-critical vehicle systems meet defined performance standards. IBA’s sub-2ms validation and human override protection provide the deterministic safety guarantees NHTSA’s AV safety framework requires.
FMVSS · NHTSA FRAMEWORK · HUMAN OVERRIDE
NIST-2025-0035
IBA is formally submitted to the federal AI agent authorization standard. AV deployments using IBA are aligned with mandatory AI authorization requirements before they are enforced for autonomous vehicle systems.
NIST-2025-0035 · mls-ubpf-pryy
SUB-2ms VALIDATION
At 100km/h, a vehicle travels 2.8cm per millisecond. IBA validates every manoeuvre in under 2ms — providing meaningful safety intervention time even at highway speeds. Safety-critical authorization at the speed the road requires.
<2ms · HIGHWAY SPEED · REAL-TIME SAFETY
PATENT GB2603013.0
The safety parameter enforcement function, hazard trajectory evaluation, and entropy-based temporal decay are protected IP. Reference implementation is Apache 2.0. Validated by xAI’s Grok as model-agnostic and straightforward to integrate.
GB2603013.0 · APACHE 2.0 · XAI VALIDATED
AUTONOMOUS VEHICLES THAT CANNOT ACT OUTSIDE SAFETY PARAMETERS
THE ROAD
DEMANDS CERTAINTY.
IBA makes AV safety authorization mathematically deterministic — not statistically probable. One conversation with the engineering team. Zero risk to the safety architecture.