Core.
The longer-term model research programme. Provider-independent development, reviewed data and evaluations come before a model release.
Research programme
Humanity.First.
Meet Max. A humanoid companion.
Designed to help. And to look out for you.
Max by Marred AIConcept render. In development.
Meet Max.
In a few moments.
Watch Max in motion Design.
Every detail, considered.
Distinctly Max.
Human proportions. A signature black visor.
A calm, approachable presence.
The visor
A rounded, uninterrupted surface with two restrained light bars. A clear identity without imitating a human face.
The body
A humanoid form with integrated armour and articulated joints. Equipment and components are being designed for easy replacement. Final dimensions, materials and mechanical specifications remain in development.
The hands
Our concept explores an open-hand gesture and human-scale proportions. Dexterity, payload and interaction limits are not yet specified.
The finishes
Pearl and Graphite are concept finish preferences. Final production colours and materials may change.
Generated design renders. Final hardware, materials and dimensions have not been announced.
Performance.
Purpose before power.
The ambition is capable intelligence in a physical form. The foundation is a system that knows where its authority ends.
Guardian.
A specialised model direction for reviewing proposed actions and their evidence. It cannot override a deterministic denial.
Model development plannedControl.
An independent action-control system. Exact authority, one-operation approvals, dispatch checks and durable evidence.
Local software prototypeWhat has been built so far?
The first local software prototype exercises bounded actions in a synthetic CRM. It includes deterministic policy checks, exact reviews, grant revocation and reconciliation of uncertain outcomes. It is a software test environment, not a deployed robot or proof of general AI safety.
Max’s compute platform, movement performance and hardware benchmarks are unannounced.
AI.
On humanity’s side.
AI should expand what humanity can do.
And humanity should remain in control.
Marred AI is building an AI defence system against misaligned AI. Max is our vision for giving that mission a physical presence: a humanoid robot designed around people, accountability and human authority.
Designed to stand
with humanity.
A calm presence, shaped around people. Familiar in proportion. Distinctly Max.
Protection is
the purpose.
Max embodies Marred AI’s mission: to defend humanity against misaligned AI.
Human control.
Always.
Our vision keeps authority with people—including the ability to stop, correct and replace an AI system.
Concept design. These principles guide development; they are not a claim of verified robot performance.
Battery.
Part of the bigger picture.
Useful movement depends on energy, thermal management and safe charging working together.
These are engineering questions for Max’s development. We will evaluate the power system alongside the physical platform, rather than promise a runtime before the hardware is validated.
- Runtime
- To be validated
- Charging
- To be specified
- Battery system
- In development
Software.
Permission comes first.
A request should become a bounded action—not an open-ended licence to act.
A clear request.One record.
One record.
One permitted change.
In the software prototype, a structured task defines exactly what an actor may propose. A request alone does not create authority.
Scope A single synthetic record
Authority Created through the trusted control path
Next step Review the exact proposed action
Illustration of the local control-software workflow. Max’s operating software and consumer interface have not been released.
Max + you.
Keep the human in control.
You decide.
A proposed action is ready for your review.
Exact actionReview before execution.AuthorityBounded. Revocable.
Review proposal
Pause new actions
Interface conceptUnderstand.
Approve.
Stop.
Our direction is a relationship where people can see what is proposed, understand the scope and retain the ability to stop or correct an AI system.
A companion app, supported devices and remote robot-control features have not been announced.
Applications.
A mission with many possibilities.
Max is a platform vision. These areas describe what we want to explore, not features available for purchase today.
AI defence.
Explore how independent action controls and specialised review could support accountable AI systems.
Physical assistance.
Investigate useful human-scale tasks, with capability limits established through hardware testing.
Human collaboration.
Develop interactions that make intent understandable and preserve meaningful human direction.
Research.
Study perception, movement, control and evaluation as parts of one system.
Interface.
Clarity you can see.
Distinctly Max.
Designed around people.
The visor and blue chest light give Max a recognisable visual language. We want its future interaction design to communicate intent and state clearly. The final indicator meanings, displays and controls remain to be defined.
Camera and sensing.
Context before action.
A robot has to understand more than a command. It has to understand the space in which it is acting.
Perception with limits.
Our research direction includes environmental awareness, uncertainty and the decision to stop when a safe action cannot be established.
Privacy by consideration.
Capture, retention and access need explicit design decisions. Camera specifications, sensor selection and recording behaviour are unannounced.
Sensing illustration. It does not depict a selected sensor array, measured field of view or demonstrated obstacle avoidance.
Connectivity.
Clear connections. Clear boundaries.
Connecting to a system should not mean inheriting unlimited access to it.
Propose
Check authority
Execute a registered action
Explicit operations.
The software prototype uses registered connector operations and bounded payloads. The actor does not receive the connector credential.
Hardware still to come.
Wireless standards, physical ports, charging connections and supported accessories for Max have not been announced.
Security.
Intelligence needs
independent boundaries.
Our software work begins with a simple principle: intelligence can propose an action. It cannot grant itself permission.
Permission.
Before action.
We’re developing controls that check a proposed action against explicit authority, independently of the AI making the proposal.
Evidence.
Beyond promises.
Proposals, approvals and observed outcomes should remain distinct, so people can see what was authorised and what actually happened.
Your control.
By design.
Our mission includes effective human control: the ability to pause, revoke, correct and replace an AI system.
These are engineering principles guiding Marred’s control software and Max’s development. They are not a claim of proven robot capabilities or universal protection.
An ambitious vision.
An honest beginning.
We’re building the foundations of Marred AI now. The first software prototype tests bounded actions, exact approvals and human stop controls in a synthetic environment.
Max is the next physical expression of that ambition. The renders show our design direction. Hardware and capabilities remain in development. Estimated delivery is 2030–2032; this is a target, not a guaranteed date.
Explore the Max concept01Build the control foundationSoftware prototype
02Develop MaxRobot design and engineering
03Test. Learn. Earn trust.Validation before capability claims
A mission this important
needs people behind it.
Max is how we intend to help fund the work.
Explore your Max. Be part of what comes next.
Standard.
US$200A standard queue position. Fully refundable.
Priority.
US$1,000A priority queue position. Non-refundable, except where required by applicable law.
Know where you stand.
Both options are non-transferable. Estimated delivery is 2030–2032, without a guaranteed date. The full robot price is unannounced.
Read reservation terms ›Review your reservation terms and continue securely with Stripe. Reservation fee only; the full robot price is unannounced.
Explore Marred.
One mission. A wider programme.
Our environment.
Part of humanity’s future.
Protecting a human future also means asking what we build, what it consumes and what happens at the end of its life.
Materials.
Material selection and sourcing will be evaluated as the hardware develops. No recycled-content percentage or environmental certification is claimed.
Energy.
Power use, charging and manufacturing impacts need measurement. Max does not yet have a published energy or emissions assessment.
Useful life.
Max is being designed for a 15-year service life, supported by easily replaceable equipment and components. Our planned refurbishment programme will bring every Max back to our factories periodically for repairs and updates, helping extend its useful life with each service. Service intervals and programme details will be confirmed as development progresses.