Mimic Labs AI Ethics Policy

Version 1.0
Effective: August 2026

Purpose

Artificial intelligence is an incredibly useful tool, but usefulness alone does not justify every possible use of it.

Mimic Labs uses AI as a tool for research, learning, analysis, auditing, troubleshooting, and communication. We do not use AI as a replacement for human creativity, technical understanding, engineering judgment, or responsibility.

Our position is not that AI should be rejected. We believe it should be used deliberately and within clearly defined boundaries.

Mimic Labs was founded by an artist, engineer, programmer, and maker. Those disciplines directly influence how we approach artificial intelligence. Creative work deserves respect for its authorship. Engineering requires understanding and accountability. Powerful tools should be approached with caution, respect, and awareness of how their use may affect others.

Our use of AI is therefore built around three principles:

Understand what you build.

AI is an auditor, not an author.

AI advises. Humans act.

These principles form the foundation of this policy.

1. The Principle of Understanding

Mimic Labs follows a simple rule:

If you work on something, you should be able to describe its function in layman's terms to anyone. If you can't, you cannot implement it into a system.

This rule is not specific to artificial intelligence.

It applies to software, circuits, mechanical systems, components, algorithms, libraries, reference designs, manufacturing processes, AI-assisted solutions, and other technologies incorporated into our work.

Using something is not the same as understanding it.

Before something becomes part of a Mimic Labs project or product, the people responsible for that work must understand what it does, why it is being used, how it interacts with the surrounding system, and the consequences of its failure.

This does not mean every person must understand every implementation detail of every discipline. Complex engineering requires specialization. It means complexity cannot be used to hide a lack of understanding.

A mechanical engineer does not need to understand every instruction executed by a microcontroller. A firmware engineer does not need to understand every process involved in producing an aluminum alloy. Both should understand how those systems affect their work, what assumptions they are relying upon, and where the boundaries of their knowledge exist.

Saying "I don't know" is acceptable.

Implementing something while pretending otherwise is not.

2. Understanding > Deadlines

Deadlines, budgets, convenience, and competitive pressure do not override the Principle of Understanding.

Mimic Labs would rather delay a feature or product than ship something its creators cannot adequately explain, maintain, troubleshoot, or take responsibility for.

This principle also applies when automation could make development faster.

AI and autonomous development tools can significantly reduce the amount of time required to complete some tasks. That reduction in time can come at the cost of removing people from portions of the investigation, implementation, and problem-solving process.

We do not consider that trade universally worthwhile.

Development time can be recovered. Understanding cannot be treated as expendable.

Mimic Labs intends to foster an environment where questions are encouraged, teaching is part of engineering, and admitting uncertainty is preferable to hiding it.

When one person understands something another does not, the preferred solution is not simply to hand over an answer. It is to communicate that understanding.

Knowledge should spread throughout a team rather than becoming isolated within individuals, departments, tools, or AI systems.

3. AI Is an Auditor, Not an Author

Mimic Labs treats artificial intelligence primarily as an analytical and advisory tool.

AI may research, observe, analyze, audit, compare, explain, suggest, and identify.

Humans create, decide, implement, and approve.

AI can help identify an error in a circuit. It can suggest areas of code that deserve investigation. It can compare components, locate documentation, explain an unfamiliar concept, challenge an assumption, or point out inconsistencies in work created by a person.

It does not become the creator of that work.

This distinction applies across engineering, software, art, design, documentation, and other work performed by Mimic Labs.

AI should help people become better at what they do rather than remove them from the act of doing it.

4. Appropriate Uses of AI

Mimic Labs permits AI to operate within a limited set of supporting roles.

These generally include:

• Research and discovery: Locating components, documentation, standards, datasheets, technical resources, and other information for human review.

• Analysis and auditing: Reviewing human-created designs, code, documentation, or other work to identify potential errors, inconsistencies, alternatives, or areas requiring further investigation.

• Education and troubleshooting: Explaining concepts, assisting with diagnosis, providing hints, and directing people toward resources that help them develop their own understanding.

• Communication assistance: Reviewing human-authored communication and documentation for clarity, accuracy, consistency, or errors.

This list is illustrative rather than exhaustive. New applications should be evaluated against the principles of this policy rather than assumed to be acceptable simply because they are not specifically prohibited.

The common requirement is straightforward:

AI assists the person performing the work. It does not perform the work in their place.

5. No Autonomous Action

AI systems are not permitted to independently act upon Mimic Labs products, projects, infrastructure, or production work.

AI may identify a problem.

AI may explain why it believes a problem exists.

AI may suggest possible approaches.

The responsible person must investigate, understand, decide, and implement.

AI systems, including autonomous or semi-autonomous agents, may be used to observe systems, analyze information, assist with debugging, run within controlled evaluation environments, or flag potential problems.

They are not permitted to independently modify production code, engineering designs, creative work, documentation, infrastructure, or other project assets.

Human oversight does not mean simply approving work already performed by an AI system.

An approval button is not a substitute for understanding.

The intended process is:

AI identifies → AI explains or suggests → Human investigates → Human understands → Human decides → Human implements

AI is granted advisory authority, not operational authority.

6. Software and Engineering

AI may assist engineers and developers with understanding documentation, locating relevant APIs, interpreting errors, diagnosing problems, reviewing human-written code, comparing implementation strategies, or learning unfamiliar systems.

AI may also be used to audit code and identify potential bugs, security concerns, unexpected interactions, or other areas deserving human investigation.

AI does not implement production code on behalf of Mimic Labs.

Autonomous coding agents are not permitted to independently write, modify, refactor, or deploy production software.

Code incorporated into a Mimic Labs project must be understood, implemented, reviewed, and tested by the people responsible for it.

This requirement is not unique to AI-generated material.

Code copied without understanding from documentation, forums, tutorials, open-source projects, examples, or any other source violates the same principle.

"The AI wrote it and it works" is not an acceptable engineering justification.

Neither is:

"I copied it from somewhere and it works."

The source is irrelevant.

If we implement it, we are responsible for understanding it.

7. Creative Work and Human Authorship

Mimic Labs does not use generative AI to create visual work.

This includes images, illustrations, concept art, product imagery, posters, graphics, textures, 3D models, renders, animation, video, branding, or other visual assets.

This applies to both public-facing and internal creative work.

AI may analyze or audit work created by a person. It may identify inconsistencies, technical problems, potential errors, or areas that deserve further attention.

It does not create or modify the creative work itself.

Mimic Labs was founded in part by an artist, and we consider human creativity to have value beyond the speed at which a finished asset can be produced.

The ability to generate an approximation of creative work does not, by itself, justify replacing the person who would otherwise create it.

Our position is intentionally simple:

AI may critique the canvas. It does not hold the brush.

8. Data, Privacy, and Intellectual Property

Mimic Labs maintains strict boundaries between artificial intelligence systems, customer information, and internal company information.

Customer Information

AI systems are not permitted access to customer information.

Customer information will not be provided to AI systems for analysis, training, personalization, research, development, or other purposes.

The appropriate level of AI access to customer information is zero.

Internal Information

Internal Mimic Labs information is not to be provided to external AI systems.

This includes unreleased designs, internal engineering documentation, proprietary source material, non-public intellectual property, internal communications, and other information not intended for external disclosure.

During the early development of Mimic Labs, external AI systems may be used where internal infrastructure does not yet exist. Information provided to those systems must therefore be limited to information appropriate for external disclosure.

As Mimic Labs develops its own facilities and computing infrastructure, our intention is to move internal AI-assisted work onto systems controlled by Mimic Labs.

Internal AI Access

Operating an AI system internally does not grant that system unrestricted access to company information.

Internal AI systems may only access resources deliberately approved for their intended function.

Access to additional repositories, files, systems, or information must be intentionally granted rather than assumed.

Self-hosted does not mean unrestricted.

AI systems should operate within clearly defined boundaries rather than being given broad access and trusted to determine those boundaries themselves.

Customer information remains customer information. Internal information remains internal information.

9. Responsible Resource Use

Artificial intelligence has a physical cost.

AI workloads consume computing resources, electricity, hardware capacity, cooling, infrastructure, and other resources whose consumption may be largely invisible to the person making a request.

Mimic Labs therefore encourages proportional use of AI.

AI should be used when it provides meaningful value to a task.

When documentation, conventional search, deterministic software, calculators, existing engineering tools, or direct human reasoning are better suited to a problem, those tools should remain the preferred solution.

We do not establish arbitrary limits based on measurements such as token counts because those measurements do not necessarily represent the actual resources consumed by different models, hardware, or infrastructure.

The goal is simpler:

Use the appropriate tool for the problem, and do not consume resources simply because they are available.

10. Transparency

Mimic Labs will not intentionally represent AI-generated work as human-created work.

AI involvement should be disclosed when its role is substantial enough that knowing about that involvement would reasonably matter to customers, collaborators, contributors, or the public.

At the same time, using AI to locate documentation, audit work, assist with debugging, explain concepts, organize information, or provide other advisory assistance does not make AI the author of the resulting work.

The distinction is whether AI assisted a person in performing their work or performed that work in their place.

Under this policy, the latter is not an acceptable role for AI at Mimic Labs.

11. Human Accountability

Responsibility cannot be delegated to artificial intelligence.

An AI system cannot accept responsibility for a defective product, unsafe design, incorrect calculation, security vulnerability, licensing violation, failed component, or harmful decision.

People can.

Responsibility for Mimic Labs products and decisions therefore remains with the humans who design, implement, review, manufacture, approve, and release them.

AI assistance does not reduce that responsibility.

No employee, contractor, collaborator, or other person working on behalf of Mimic Labs may use artificial intelligence as justification for a decision they cannot independently understand and defend.

If we choose to trust a tool, we remain responsible for that choice.

12. A Living Policy

Artificial intelligence is changing rapidly.

Its capabilities, limitations, resource requirements, training practices, legal environment, and relationship with creative and technical work will continue to change.

This policy is therefore intended to evolve alongside our understanding of the technology.

Restrictions may become stronger where new risks emerge. Existing restrictions may be reconsidered where technological or industry changes meaningfully resolve the concerns that created them.

Changes to this policy should be driven by evidence, experience, and careful consideration rather than convenience or technological enthusiasm.

Until a change is deliberately made, the boundaries established by the current version of this policy remain in effect.

Core Principles

Mimic Labs does not reject artificial intelligence.

We reject the idea that every task should be automated simply because automation is possible.

AI is extraordinarily useful for learning, research, auditing, troubleshooting, and challenging our assumptions. Used responsibly, it can help people understand more and make better-informed decisions.

That usefulness does not require surrendering authorship, responsibility, privacy, or understanding.

Our approach can therefore be summarized in three principles:

Understand what you build.

If you cannot explain something well enough to communicate its function in simple terms, you are not ready to implement it.

AI is an auditor, not an author.

AI may research, analyze, compare, explain, audit, and suggest. Humans create and implement.

AI advises. Humans act.

AI is not granted independent authority over our products, projects, infrastructure, or creative work.

Above all:

If we build it, we are responsible for understanding it.

Revision History

Version 1.0 | August 2026
Initial publication of the Mimic Labs AI Ethics Policy.