A person has an identity
The identity is generated locally and can be used to sign data, authenticate and communicate.
MAX
Humans, Machines & AI Agents
What if humans and machines used the same trust model from the beginning?
A person has an identity. A machine has an identity. The person can sign rules. The machine can verify them.
MAX explores a common architectural model for identity, signatures, authorization and communication across people and machines. AI agents can participate without becoming the source of authority.
The idea is simple
The individual technologies already exist. The question is whether there is value in making them part of one coherent trust architecture from the start.
A machine has an identity
A device can also have its own cryptographic identity and participate as an identifiable entity.
The machine can verify rules
A person can sign an authorization. The machine can verify it before deciding whether to perform an action.
Person → signed authority → machine → verify → execute or reject
The individual pieces are not new
Digital identity, machine identity, signed authorization, policy enforcement and local verification already exist in many forms.
The question is the continuity between them
Existing systems can connect identity providers, device identities, policy engines, gateways and agent frameworks.
MAX explores a different starting point: people and machines are designed as participants in the same broader model of identity, signatures, communication and verification.
The architectural question
Is there practical value in using one coherent trust model from the person all the way to the machine that performs the action?
One model, different identities
“Common model” does not mean that a person and a machine share the same identity. Every participant has its own identity.
Who are you?
Humans and machines can each have their own cryptographic identity inside the same architectural family.
Who authorized this?
Signed data can provide cryptographic evidence about the origin of an instruction or authorization.
Should I act?
The system that performs the action can verify whether the requested action is actually authorized.
The server can coordinate without becoming the identity
Central infrastructure can still be useful without needing to hold the private secrets from which the identity is derived.
Coordination and authority are different roles
In the MAX model, identity material is generated locally. The server can coordinate communications and maintain the public associations needed by the system.
But the final system can still verify signed information itself before accepting or executing an action.
A useful separation
The infrastructure can help participants communicate. The endpoint can still make its own verification decision.
Then add an AI agent
The AI can make the interaction more natural and powerful, without becoming the root of trust.
The agent can ask. The machine verifies.
An AI agent can interpret a person's request, reason about it, coordinate work and send a request to a machine.
But the machine does not need to accept the action merely because the AI requested it. It can verify whether the request remains inside the authority previously granted by the person.
This means the model does not need to depend on one specific AI provider, model or agent framework.
The important boundary
The intelligence can belong to the AI. The authority does not have to.
A simple example
Imagine a home controller with its own machine identity.
Define the limits
The person signs rules describing what the controller is allowed to do and under which conditions.
Request an action
Any compatible AI agent can receive a natural-language request and ask the controller to perform an action.
Verify before acting
The controller checks the signed authority. If the action is permitted, it can execute it. If not, it rejects it.
A manipulated or malfunctioning agent may still make bad decisions inside the authority it legitimately has.
But compromising the agent alone should not automatically allow it to expand that authority if the machine-side verification remains intact.
This is not starting from a blank page
The human and machine sides already exist as working software. The complete AI-agent flow is the next area of experimentation.
MAX App
MAX App is available for iPhone and explores local digital identity, signatures, authentication, protected data and encrypted communication.
MAX IoT
MAX IoT extends the model to physical devices. The current prototype runs on Raspberry Pi devices with machine identity, communication components and signed administrative manifests.
Post-quantum components
The prototype already uses post-quantum cryptographic components, including SPHINCS+ and FrodoKEM.
Prototype status
MAX IoT is an experimental prototype, not an industrially certified system. Security verification and hardening are ongoing. The complete human → AI agent → machine authorization flow is not being presented as a finished protocol.
Why could this be useful?
The potential value is not a new cryptographic primitive. It is the possibility of reducing the number of separate trust layers that must be connected later.
One architectural direction
Human identity, machine identity, signed authority and verification are explored as parts of one coherent model.
Verification close to the action
Important authorization decisions can remain close to the device that actually performs the physical or operational action.
The AI can be replaceable
The trust relationship does not have to be rebuilt every time the AI model, provider or agent framework changes.
If this architecture can make human-to-machine delegation simpler to integrate, easier to verify and less dependent on a single provider, it could become useful well beyond one application.
The question I am exploring
I am not claiming that identity, authorization, machine verification or agent governance are new concepts. They are not.
Can the combination become useful?
Can one coherent trust architecture connect people, machines and AI agents with fewer integration layers?
Can the final machine retain the ability to verify what it is actually authorized to do?
Can AI remain powerful and replaceable without becoming the source of human authority?
For engineers, researchers and builders
I want to understand where this model overlaps with existing architectures, where it differs, where it could interoperate with them — and where it breaks.
The agent can ask. The machine verifies.
MAX is exploring whether humans, machines and AI agents can interact through one coherent trust architecture without moving human authority inside the AI itself.