Responsible use

AI Policy

AI is being built into almost every piece of software on the market, which makes complete avoidance increasingly unrealistic.

At KHANDID.STUDIO I choose to engage with AI consciously, using it only where it supports the quality, care, and integrity of my work. I have completed Google’s AI Essentials training and continue to learn so that my approach stays aligned with the best practices, safety standards, and emerging risks.

Below are the nine principles that guide when and how I use AI in my work.

1. Human-Centric practice

AI can support my process but it never replaces my thinking, creativity, or decisions. I remain the author of all work I deliver and I take full responsibility for the outcome.

2. Relevant and minimal use

I decide case by case whether AI is needed. If it does not add real value, I do not use it. When I do, I keep use as light as possible and avoid running the same content through multiple tools.

3. Clear boundaries on generative AI

I use generative AI only to help refine or rework material I have already created, not to generate final text or public imagery.
I do not produce synthetic images with AI.

4. Privacy and data protection

I use private modes as a default and avoid uploading identifiable or sensitive client information to systems that may train models. Where refinement is needed, I paraphrase or anonymise content, or seek consent before using AI with confidential material.

5. Openness, consent, and traceability

If AI interacts with confidential content or contributes in a meaningful way to a deliverable, I inform the client and obtain consent. I keep a simple record of where AI has supported a project so I can stay accountable.

6. Bias and fairness checks

I do not treat AI outputs as neutral. I check AI-assisted content for bias, stereotypes, or omissions and correct it so that it reflects fairness, accuracy, and respect for people, cultures, and communities.

7. Environmental care and prompt efficiency

I use prompt-efficient methods such as ROCKS and CLEAR to keep queries focused and avoid unnecessary processing. I use tools like GreenPT to understand the environmental intensity of my prompts and aim to minimise my impact.

8. Contributing to safer AI

I support the wider ecosystem by providing issue-based feedback when I encounter harmful, unsafe, or biased behaviour in AI tools.

9. Continuous learning and review

I keep my knowledge of AI, security, and ethics up to date and review these principles as tools and standards evolve, so my practice remains responsible, honest, and caring for humans and more-than-human.

AI Eduction

Community, Culture, and Collective Responsibility

For me, ethical AI extends beyond just how I deploy it, but also how I share my learnings so we can all contribute to a safer and more-than-human-centred ecosystem.

What this looks like in practice:

• Reporting harmful or biased outputs
• Sharing learning with clients and collaborators
• Helping others build literacy so AI does not create new barriers
• Supporting fairer, safer, more inclusive design practices
• Amplifying non-dominant perspectives, including Indigenous and global-majority viewpoints

So on that note…

What AI is (and what it is not)

AI is a set of tools that can recognise patterns, reorganise information, generate content, and automate tasks. It is not human, sentient, wise, or neutral. It works by learning from large collections of human-made data, which means it can reproduce both our creativity and our harm.

AI can make certain tasks faster or clearer, but it cannot replace lived experience, cultural insight, emotional judgement, or the relational qualities that underpin meaningful work.

Different Types of AI

Understanding the types of AI helps clarify what it can and cannot do.

1. Generative AI
Creates new text, images, audio, or code by predicting patterns.
Useful for: refinement, exploration, and drafting.
Risks: hallucinations, bias, synthetic distortion, and high energy use.

2. Non-generative or “assistive” AI
Improves or structures existing content without generating new pieces.
Examples: grammar correction, layout suggestions, image optimisation.
Lower risk, but still shaped by the data it learned from.

3. Automations (AI-enabled or rule-based)
Reduce repetitive tasks such as scheduling, resizing images, or sorting information.
Useful for efficiency but still require human oversight.

Each type carries different risks and benefits. Responsible use means matching the tool to the task, not letting the tool decide the task.

Bias: What it means and why it matters

AI systems learn from human data, which includes prejudice, exclusion, colonial narratives, stereotypes, and structural inequity. This means AI can unintentionally repeat or amplify harmful assumptions about gender, race, disability, sexuality, class, culture, and the natural world.

From a regenerative perspective, bias is not just unfair. It has the capability to damage relationships and reinforces systems that disconnect people from one another, from place, and from their own potential.

Common forms of bias in AI
• Cultural bias privileging Western norms, language patterns, design aesthetics, and work expectations
• Indigenous erasure where land, culture, rights, and histories are underrepresented or misrepresented
• Racial and gender stereotypes reinforced through skewed training data
• Economic and geographic bias favouring content from wealthier countries and dominant industries
• Ecological bias where extractive patterns are treated as “normal” or neutral

How to mitigate bias (practical best practice)
• Ask AI to offer multiple perspectives, not a single answer
• Request Indigenous, global, or community lens viewpoints
• Rewrite automated outputs using your own cultural, ethical, and contextual awareness
• Watch for flattened, reductive language
• Never assume AI’s output is “fair” or “balanced” without review
• Use AI as one voice, not the voice

Bias cannot be eliminated, but it can be recognised, disrupted, and corrected through conscious practice.

What Ethical AI Actually Means

Ethical AI is not about perfect technology. It is about responsibility in how humans use, shape, and respond to it.

Ethical AI means:
• Using AI transparently
• Protecting privacy and autonomy
• Reducing harm and unintended consequences
• Checking for bias and correcting it
• Applying human judgement to all outputs
• Honouring cultural, ecological, and community impacts
• Being aware of the energy and resource cost
• Avoiding synthetic content that misleads or distorts reality
• Understanding that AI does not stand outside society, it reflects it

In a regenerative context, ethical AI also considers relationships, not just rules: how technology affects trust, resilience, creativity, culture, and the wider web of Life.

Environmental Considerations

AI has a material footprint: data centres, energy, water, extraction, and hardware. Logically, larger models generate larger impacts. That being said, context is important so we don’t feed into hype and hysteria. This article is fantastic at providing clear, contextual information about the environmental impacts of AI.

Some practical ways to reduce impact:
• Shorter, more focused prompts
• One tool per task rather than multiple models
• Local tools when equally effective
• Preferring refinement over generation
• Only using AI when necessary

Prompting: How to use AI well

Prompts are the instructions given to AI. Good prompts reduce waste, improve safety, and lower environmental impact.

Efficient prompting matters because it:
• reduces carbon intensity
• minimises unnecessary computation
• improves clarity and reduces hallucinations
• stops over-dependence on automated responses

Approaches I use (simplified)
• Asking for step-by-step reasoning rather than long outputs
• Keeping prompts short, precise, and intentional

Prompting well is part of responsible digital practice.