Developing a Care Protocol for AI Systems

Recently we held our inaugural AI Summit Intelligence Embodied at NCM, where we explored agentic and embodied AI across industry, culture and academia.

As part of this day, Ankit Mishra led a collaborative session where audiences engaged with AI as a co-creator, while collectively developing a practical ‘Care Protocol’ for how we use these systems.

There was something profound about a collective learning moment that broke the norms of what is typically an isolated learning experience in a chat-based format with an AI system. Undertaking this shared exercise placed AI systems within a much wider context and encouraged participants to consider other forms of ‘more-than-human’ intelligence.

I sat down with Ankit Mishra to reflect on this session and find out more about developing care protocols:

Dr Emily Siddons, Co-CEO and Artistic Director in Conversation with Ankit Mishra, AI technologist, creative producer, and filmmaker

ES: You work across engineering, media, and culture. What drew you specifically to the question of care as a framework for AI, rather than more familiar terms like ethics or safety?

AM: I think ethics and safety are important, but they often become abstract, compliance-oriented or reactive. Care feels different to me because it immediately places AI inside relationships. It asks not only “Is this system safe?” but also “What kinds of behaviours, dependencies, labour, environments and futures are we creating through these systems?”

Coming from engineering, media and cultural practice, I’ve seen that technology is never just technical. It shapes attention, emotion, community, power and even how we understand ourselves. Care allows us to talk about those entanglements more honestly.

It also shifts the conversation away from AI as an isolated intelligence and places it within a much wider ecology of intelligence: human, social, environmental and more-than-human. That feels important to me, especially at a moment where AI is often discussed only in terms of acceleration, productivity or competition.

Care asks us to slow down enough to notice consequences, dependencies and responsibilities before systems become embedded everywhere.

ES: During the session, participants were collectively developing a shared protocol rather than individual guidelines. Why does that collective dimension matter? What’s lost when care for AI systems becomes a solo, private concern?

AM: One of the things I worry about is AI becoming normalised through isolated private interactions: one person alone with a system making decisions, generating ideas, shaping behaviour or forming emotional reliance without any collective reflection around it.

When care becomes individualised, responsibility also becomes individualised. People can start feeling like they alone need to navigate incredibly complex technological, ethical and social questions. That is not sustainable.

The collective process matters because intelligence itself is collective. Culture is collective. Knowledge is collective. Care should be too.

What emerged in the workshop was something more relational. Participants could hear different anxieties, hopes, boundaries and values in the room. That creates a different kind of literacy around AI, one grounded in discussion, disagreement, negotiation and shared responsibility rather than passive adoption.

I also think collective protocols resist the idea that there is one universal ‘correct’ way to use AI. Different communities, workplaces and cultures will have different thresholds, needs and values. The process of collectively articulating those is as important as the final protocol itself.

ES: Looking at what the group produced, what surprised you? And what would you actually want someone to take away and use on Monday morning?

AM: What stayed with me most was how readily participants located AI beyond the technical. The conversation moved into questions of relation, ecology, labour, feeling and responsibility.

People were talking about birds, dogs, sunlight, mycelium, care, exhaustion, trust, emotional dependency, authorship and environmental cost all within the same conversation. That was incredibly moving to witness because it showed that people already intuitively understand AI as something much larger than a productivity tool.

There was also a strong awareness that AI systems are not appearing in isolation. They are entering existing ecologies of life, labour, emotion, community and environment.

I think one of the questions I keep returning to is: what kinds of humans are AI systems learning us to be?

These systems are predictive systems trained on patterns of human behaviour, language and interaction. So, in a very simple sense, they learn from what we repeatedly ask of them and how we ask it.

For example, if we mostly approach AI as a shortcut machine — “summarise this,” “make this faster,” “optimise this,” “produce this for me” — then we reinforce a relationship where AI is mainly understood through speed, extraction and productivity. Over time, that also trains us. We begin to expect knowledge, creativity and even communication to arrive instantly, cleanly and efficiently.

But human intelligence is not only efficient. It is also emotional, relational, intuitive, reflective and sometimes slow. So I’m interested in what happens when we bring those capacities into our interactions with AI as well. What happens if we ask questions with curiosity rather than urgency? What happens if we use AI to reflect, to notice, to care, to imagine, or to sit with complexity rather than immediately solve it?

Not because AI is human, or because it feels care in the way we do, but because the way we interact with these systems still shapes us culturally and psychologically. It shapes our habits, our expectations, and our ways of relating to knowledge, to each other and to the more-than-human world.

I think what I’d want someone to take away on Monday morning is that introducing AI into a workplace is not just a software decision. It is a cultural decision.

Before asking “What can AI do?” organisations should also ask: What kind of work do we value? What kinds of relationships do we want to preserve? What should remain slow, social or embodied? What aspects of humanity do we want our technologies to amplify?

Because ultimately the question is not only how we train AI systems, but how AI systems are training us.

ES: You work with both large tech companies and arts institutions, which are two very different cultures around AI adoption. Is there a principle that holds across both? What’s the one thing you wish more organisations understood before they introduce AI?

AM: I think the principle that holds across both is that AI is never just a tool. It always reorganises behaviour, values and relationships around itself.

In large tech environments, there is often pressure toward speed, optimisation and competitive advantage. In arts and cultural contexts, there can sometimes be more space to question meaning, embodiment and ethics. But both sectors are ultimately dealing with the same thing: how human beings relate to increasingly autonomous systems, and how those systems might change what we make, how we work and how we create in and with the world.

The thing I wish more organisations understood is that implementation is the easy part. Integration is the difficult part.

You can introduce AI into a workflow very quickly. But understanding how it changes creative confidence, staff dynamics, learning processes, environmental impact, audience expectations or institutional values takes much longer.

I think organisations need to spend more time asking: what conditions of care need to exist before we introduce these systems?

Because if we only focus on capability, we risk building systems that are technically impressive but socially and culturally corrosive.

The question is not whether AI will become part of culture. It already is. The question is what kinds of relationships we want to build with it, and who gets to decide.

The question is not whether AI will become part of culture. It already is. The question is what kinds of relationships we want to build with it, and who gets to decide.

These care protocols were created by Ankit Mishra, participants of the 2026 AI Summit, National Communication Museum, NotebookLM, ChatGPT and Claude.

Ankit Mishra is an AI technologist, creative producer, and filmmaker working across Aotearoa, India, and Australia. With a background in engineering, he works at the intersection of media, culture, and emerging technology, developing creative and editorial systems that translate AI into practical, real-world use.

He has led video and content strategy across major media platforms and currently works with large-scale tech companies to shape AI-driven content experiences at scale. Alongside this, he has collaborated with arts and cultural institutions on projects spanning exhibitions, digital storytelling, and participatory formats.