01
Authority. Per action.
Access to a system is not a license for every action. Check which robot is authorized to do what, and under which conditions.
AUTHORITY PHYSICAL AI · AUTHORITY & EVIDENCE
Authority and Evidence Layer
for Physical AI Actions
For robot actions.
Verifiable authority.
Evidence you can check.
An AI command is not permission to act.
PAIL checks action permissions and enables
verification of authorization and decision records.
THE NEXT QUESTION FOR PHYSICAL AI IS AUTHORITY.
AI-generated concept · Not a real product or customer deployment
AUTHORITY FIRST.Evidence you can verify.
✓01 / THE LAYER BETWEENExplore PAIL ↓
A command is not
the same as
permission to act.
AIAGISI
As intelligence advances,
authority must become clearer.
In an era discussing AI, AGI and SI, physical actions still need authority and evidence. PAIL checks action permissions in the integrated execution path.
How we use AI, AGI and SI.
AI means Artificial Intelligence, AGI means Artificial General Intelligence, and SI means Super Intelligence. The U.S. executive-branch terminology change is not a declaration that technical superintelligence has been achieved. This is not a claim that PAIL has been integration-tested with every AGI or SI system.
Official U.S. terminology notice ↗02
More than an alert.
PAIL does not just flag unauthorized commands. It prevents them from being forwarded through the integrated execution path.
ENFORCEMENT03
Records you can verify.
Verify the authenticity and integrity of authorization and decision records, so they can support operational review.
EVIDENCEConnect PAIL to
your robot system.
Integrate through an adapter in the command path, rather than replacing your AI and robot controller wholesale. Defined interfaces and reference implementations let you start with one task scenario.
ROS 2 REFERENCE · SOFTWARE EVALUATION- 01
Identify the integration point.
Review how the robot receives commands and how its existing control path works. Define where PAIL fits.
- 02
Start from a reference.
Explore authorization with the ROS 2 reference path and prepared software examples. You do not have to build every component from scratch.
- 03
Verify on your robot.
Prepare the adapter and deployment conditions for your environment, then check the integration against defined cases.
Integration scope and effort depend on robot interfaces and operating conditions. We do not promise instant support for every model.
Explore PAIL for our robot ↗03 — VERIFY IT YOURSELFDon’t just take our word.
Don’t just take our word.
Check it yourself.
Use evaluation examples and verification procedures to check command authorization, refusal and record verification. Start in a prepared software environment, then move to integration tests on your robot.
Request evaluation guidance↗Tell us your robot and the task you want to evaluate. We will outline the materials and environment needed.
Four things to check01—04
- 01REFUSAL
A command without permission
Check that the command is refused and not forwarded through the integrated path.
- 02FORWARDING
A properly authorized request
Check that the permitted command reaches the test output.
- 03BOUNDS
A request outside permitted bounds
Check that a command exceeding defined conditions is refused with an explicit reason.
- 04VERIFICATION
Records used for review
Check authenticity and integrity using the supplied records and verification material.
These are evaluation cases, not results from a test run on this page. Software tests do not replace physical stopping-performance tests or safety certification. Record verification requires the corresponding verification material.
PAIL / CONCEPT FILMNarrated · 41 seconds
PAIL / THE NEXT ACTIONAuthority to act.Evidence to verify.Authority and evidence. For robot actions.▶
04 — WHY NOW
EU MACHINERY REGULATION2027JANUARY 20
General application begins
(EU) 2023/1230
For teams preparing for the EU market
Preparing robots for Europe?
Prepare their authority
and evidence, too.
The EU Machinery Regulation addresses protection of safety-related software and data, cyber-safety of control systems, and AI-powered safety functions. Explore where PAIL’s action authorization and evidence can support your preparation.
- 01Which actions should be authorized?
- 02Which records can you check and verify?
- 03What else does product-level assessment need?
PAIL does not replace conformity assessment or guarantee product-level compliance, CE marking or safety certification.
European Commission guidance ↗Checked: 6 October 2026For the teams building robots.
And those operating them.
Add action authority and evidence
to the intelligence and robots you already have.
ROBOT MAKERSRobot manufacturers
Keep action permissions explicit when connecting external AI. Explore your command path and integration options with us.
Explore an integration ↗
SYSTEM INTEGRATIONSystem integrators & AI teams
Separate an AI’s action request from the conditions for execution. Start with one practical integration scenario.
Explore an integration ↗
ROBOT OPERATIONSRobot operators
Review collected robot states and decision histories. Define the records and evidence your operational review needs.
Explore an integration ↗Applicability depends on robot interfaces and deployment conditions. AI-generated images are concepts, not customer deployments or compatibility claims.
06 — START WITH ONE SCENARIOStart with
Start with
one scenario.
A clearly scoped first step, not a sweeping promise.
Understand the environment. Agree on how to evaluate.
- 01
Understand the environment.
Tell us about the robot, how it operates, and the problem to solve.
- 02
Define the scenario.
Choose the action permissions and records you want to evaluate.
- 03
Plan the evaluation.
Agree on integration scope, verification methods and additional requirements.
07 — GOOD QUESTIONSClear answers.
Clear answers.
Good starting points.
Does PAIL replace AI or the robot controller?
PAIL does not replace your AI model or robot controller.
It adds action-permission checks to the intelligence and control systems you already use. It checks which robot is authorized to perform which action under defined conditions, and stops unauthorized commands from being forwarded through the integrated execution path. Verifiable authorization and decision records give your team both execution criteria and a basis for review.
What does “verifiable evidence” mean?
Records alone do not establish every physical action, the cause of an accident, or legal liability.
They can provide a checkable basis for trusting authorization and decision records. Review the collected history of permitted and refused actions, then verify the authenticity and integrity of those records. Your team can investigate operational issues using evidence it can check, rather than relying only on assumptions.
Does it replace existing safety systems?
PAIL does not replace emergency stops, protective devices or other existing safety functions.
It complements them by taking responsibility for a different question: was this action authorized? While robot safety systems address physical hazards, PAIL checks action permissions and blocks unauthorized command forwarding. You can retain existing safety functions and add a distinct action-authorization layer.
Which robots can use PAIL?
We do not promise instant support for every robot. Integration depends on the command interface and operating environment.
Share your robot and task, and we can turn applicability into a concrete evaluation plan. We review the command path, select an initial action scenario, and agree on the integration work and checks. You can start with one scenario rather than changing an entire deployment at once.
Does adopting PAIL ensure EU compliance?
Adopting PAIL alone does not guarantee CE marking or product-level compliance.
PAIL contributes action-permission controls and verifiable records to your preparation. Define which actions to permit, and use authorization and decision records as part of operational and conformity review materials. We help distinguish the PAIL scope from risk assessment, safety functions and technical documentation that still need separate preparation.
How can our team start an evaluation?
Tell us your robot and the task to evaluate, and we will outline a starting path. Prepared software examples let you explore basic authorization without physical robot hardware. After confirming the environment and available materials, check permission, refusal, command forwarding and record verification before moving to integration tests on your robot.
LET’S DEFINE THE NEXT ACTION.
The next robot action.
With authority. With evidence.
Tell us about your robot and the task you have in mind.
Let’s find the right place to start.