Independent verification for autonomous systems

Independent ratingsfor Physical AI.

BoundaryScore™ independently measures how autonomous machines move and perform in the real world, then turns that evidence into a clear, comparable rating.

See how it works Coming in 2027

Why It Matters

Self-reported performance is not a common standard.

Physical AI companies use different metrics, definitions, and reporting practices. The result is evidence that is difficult to reconcile—and even harder to compare.

BoundaryScore™ is being developed to provide the independent external reference the market is missing.

01

Different measures

Proprietary definitions produce incompatible results.

02

Different evidence

Self-reported data is not designed for independent comparison.

03

One external reference

A consistent framework creates a credible basis for decisions.

How It Works

From a machine in motion to a trusted rating.

A measurement device is installed on the machine. Using TraceQ™, it observes real-world motion independently of the machine’s own software and reported state.

Measurement

  1. 01

    Install

    Fit an independent measurement device to the machine.

  2. 02

    Measure

    Observe real-world motion independently of onboard systems.

  3. 03

    Reconcile

    Compare observed behavior with machine-reported state.

  4. 04

    Rate

    Produce a comparable score with traceable supporting evidence.

Designed across Physical AI

One framework for machines that move.

  • Autonomous vehicles
  • Passenger and commercial vehicles
  • Bipedal robots
  • Quadruped robots
  • Mobile and industrial machines

The Rating

A result people can read—and evidence experts can examine.

A BoundaryScore rating communicates performance against a defined benchmark and operating conditions. It is supported by traceable evidence and should not be read as a standalone safety certification.

The example shown is illustrative. Final bands and thresholds will be published with the methodology.

BoundaryScore™Illustrative rating
87/100Strong
Evidence status
Independently verified
Operating domain
Defined conditions
Supporting record
Traceable evidence

Built for decisions in underwriting, deployment, operations, investment, and procurement.

Standards and Neutrality

Built on standards. Designed to evolve.

BoundaryScore™ works with applicable standards, regulation, and domain expertise. It is designed to extend as technologies, operating conditions, and market requirements change.

Open and extensible

Built to expand across systems, domains, and evidence types.

Standards-aligned

Complements applicable standards and regulatory frameworks.

Auditable

Inputs, calculations, and evidence remain traceable and reviewable.

Independently governed

Designed to evolve through structured industry participation.

Governance

Clear roles. Independent infrastructure.

Benchmark

BoundaryScore™

The public-facing framework for evaluating and communicating Physical AI performance.

Infrastructure

Paverly

Builds and operates the measurement, reconciliation, and scoring infrastructure.

Governance

Boundary Collective™

A non-profit industry association supporting transparent, multi-stakeholder governance.

Who It Serves

One benchmark. Better decisions across the market.

BoundaryScore™ gives stakeholders a consistent basis for evaluating performance without relying solely on incompatible or self-reported claims.

01

Oversight and public trust

Regulators, standards bodies, media, and the public gain a clearer, traceable reference.

02

Capital and risk

Insurers, financial institutions, and investors gain comparable evidence for risk and investment decisions.

Resources

The thinking behind the benchmark.

Methodology, governance and evidence for decisions across the market.

Foundational paper

Independent Verification for Physical AI

Explains the case for an external reference beyond the machine's own reporting systems.

Read whitepaper ↗

Governance Q&A

Boundary Collective™ Governance

Explains how the industry—not one company—governs scoring, methodology, data and privacy.

Read Q&A ↗

Data protection brief

BoundaryScore™ Data Covenant

Eight commitments governing how confidential participant evidence is collected, protected, used and disclosed.

Read brief ↗

Data rights Q&A

BoundaryScore™ Data Rights Q&A

Answers what evidence is required, who owns it, and who can — and cannot — see participant-specific data.

Read Q&A ↗

Interactive guide

Explore the Data Covenant

The eight commitments and the Who Can See What? tool, in one interactive page.

Explore →

Regulators & standards Q&A

A Real-Time Trust and Validation Layer

Explains how BoundaryScore™ complements oversight with continuous, affordable and industry-governed evidence.

Read Q&A ↗

AV safety brief

The Missing Independent Signal for AV Safety Leaders

Explains why AV safety teams need an independent motion signal for monitoring, validation and investigation.

Read brief ↗

Insurance Q&A

Quantifying Performance and Insurer Risk

Shows how consistent, physics-based and auditable evidence can improve risk understanding and pricing.

Read Q&A ↗

Consumer Q&A

Understanding a BoundaryScore™ Rating

Clarifies what the rating means, what it does not mean and why availability will vary by manufacturer.

Read Q&A ↗

Physical AI Q&A

Building Trust in Physical AI

Shows how independent, granular evidence supports builders, operators and the deployment ecosystem.

Read Q&A ↗

Technical Q&A

Motion Validation Envelope™

Shows how alignment creates confidence and meaningful discrepancy creates evidence for review.

Read Q&A ↗

Foundational paper

TraceQ™: A New Way to Measure Motion

Introduces direct motion measurement as the technical foundation for independent machine-performance validation.

Read whitepaper ↗

Technical Q&A

Why Motion?

Explains why motion is universal, economical, privacy-enabling and uniquely suited to independent validation.

Read Q&A ↗

1–3 of 13