AI and the Dharma of Liberation: What Would It Mean to Build Ethical Machines
There is a question that has been sitting at the centre of the AI debate for years, mostly unasked in polite company. Not whether AI is powerful. But whose power it serves.
We hear a great deal about alignment. We hear about safety, about guardrails, about responsible deployment. These are real concerns and serious people are working on them. But there is a prior question that rarely gets asked with any urgency, and that is the question of intention. Before we ask whether an AI system is safe, we might ask what it is safe for. Before we ask whether it is aligned, we might ask aligned with what, and aligned with whom.
I want to try to answer that question from an unusual angle. I want to draw on Buddhist philosophy, on the systems thinking of Buckminster Fuller, and on the materialist critique of the left. These are not natural allies. But I think they converge on something important: the idea that technology is never neutral, that the values built into a system at design time scale up to world-shaping levels, and that we are currently building systems at enormous scale without anything close to the ethical clarity that would justify doing so.
The Mirror That Amplifies
The first thing to understand about AI is that it is not a tool in the ordinary sense. A hammer does not encode a worldview. But an AI system trained on human-generated data, optimised toward human-defined objectives, and deployed in human social contexts is saturated with values at every layer. It reflects the assumptions of its designers, the biases in its training data, and the incentive structures of the institutions that fund it. And then it amplifies all of that, at speed and at scale, across millions of interactions.
This is what makes the current moment so consequential. The question of what values we embed in these systems is not an abstract philosophical puzzle. It is a question with material consequences for billions of people. A recommendation engine that optimises for engagement does not merely show you interesting content. It reshapes your attention, deepens your outrage, and slowly remodels what you believe. A hiring algorithm that penalises certain postcodes does not merely filter CVs. It perpetuates structural inequality with the appearance of objectivity. A surveillance system deployed in a poor neighbourhood does not merely observe. It criminalises poverty with algorithmic authority.
We are building mirrors, and the mirrors are making what they reflect larger.
If we’re going to talk about ethics in AI, we need to go deeper. We need to go beyond just codes of conduct or diversity checklists. Rather, we need a vision of what kind of society AI should serve. A vision that includes the insights of Buddhism, but also the bold critiques of communism, anti-work movements, and thinkers like Buckminster Fuller. This is because AI is not just a tool. It is also a mirror and an amplifier. It reflects the values of its creators and scales them up to world-shaping levels.
Planning for All Life, Not Private Profit
Buckminster Fuller, the futurist architect and systems theorist, believed in using technology to solve global problems, not just for profit, but for human thriving. (https://en.wikipedia.org/wiki/Buckminster_Fuller) He imagined a society where planning, design, and engineering were guided by one question:
“How do we make the world work for 100% of humanity in the shortest possible time, through spontaneous cooperation, without ecological damage or disadvantage to anyone?”
See more details on one of his ideas, called World Game: https://www.bfi.org/about-fuller/big-ideas/world-game/
But that is unfortunately far from the vision behind today’s AI.
Today, we have some of the most powerful models in history, but they are being trained and deployed to:
- Increase ad clicks and sell more products
- Predict and suppress dissent
- Automate layoffs and optimize shareholder value
- Help military planners target faster and surveil better
The AI systems are brilliant in their capabilities. But they are used for goals which are narrow and often destructive.
His answer was not politics in the conventional sense. He did not believe you could change the world by fighting the existing system head-on. He believed you changed it by making the existing system obsolete, by designing something better. His concept of the World Game was a serious attempt to use the best available data and systems modelling to allocate the world’s resources more intelligently than markets or nation-states were managing to do. He thought that scarcity, in most of its forms, was a design problem rather than a natural fact. He thought that intelligence applied with genuine intention toward human flourishing, rather than toward private accumulation, could solve it.
Fuller was eccentric and sometimes wrong. But he was right about something fundamental: that planning and design guided by comprehensive thinking can accomplish what competition and self-interest cannot. And he was right that the technologies of his era, computers, communications, logistics, could in principle be organised toward planetary benefit rather than private profit.
AI is exactly the kind of technology Fuller was imagining. It can model complex systems. It can coordinate resources across scales that human institutions cannot manage. It can process the kind of comprehensive data that genuine planetary planning would require. The infrastructure, technically, is nearly there.
What is not there is the intention.
Instead, today’s most powerful AI systems are being trained and deployed to increase ad clicks, optimise supply chains for shareholder value, predict and manage consumer behaviour, help military planners target more efficiently, and automate the parts of labour that are most costly to capital. These are not edge cases or minor misapplications. They are the core use cases. They are what the funding is for.
Fuller would have found this not just wrong but wasteful. A squandering of something genuinely remarkable on goals that are, by any serious measure, embarrassingly small.
Imagine if AI were used for the following tasks:
- Real-time systems for food redistribution
- Smart ecological planning to reverse climate damage
- Universal access to education, healthcare, and clean water
- Tools to amplify indigenous knowledge and collective wisdom
The infrastructure is technically possible. What’s lacking is intention.
From Buddhist Ethics
Buddhist ethics begins not with rules but with a question: what is the nature of suffering, and what perpetuates it?
The Buddha’s diagnosis was precise. Suffering arises from craving (tanha), from aversion, and from the fundamental confusion (avidya) about the nature of self and reality that makes craving seem like a reasonable strategy in the first place. The path out of suffering is not the satisfaction of craving but its dissolution. Not through suppression, but through clarity. Through seeing what is actually happening in experience rather than through the distorting lens of habitual reactivity.
What happens when we apply this diagnostic lens to AI?
Most AI today is trained on data that is, in the Buddhist sense, soaked in samsara. The internet, from which the bulk of large language model training data is drawn, is an enormous archive of human craving, aversion, and confusion. It is the outward expression of a species that is, in the aggregate, not particularly awake. Training models on this data and then amplifying their outputs at scale is a little like building a machine to mass-produce the ego and calling it progress.
More concretely: recommendation engines are explicitly optimised to exploit craving. They are not designed to give you what is good for you. They are designed to keep you engaged, and engagement, in human psychology, correlates not with satisfaction but with emotional arousal. Which means outrage, anxiety, envy, and longing. Facial recognition systems, trained on racially biased datasets, encode aversion in algorithmic form. Predictive policing reproduces the structural delusions of a society that has already decided certain people are inherently dangerous.
Buddhism does not reject tools. The tradition has always engaged with the material world, and a hammer used to build a monastery is not the same as a hammer used to break someone’s head. The question is always one of intention and effect: Is this action coming from wisdom and compassion, or from greed, hatred, and delusion? Does it reduce suffering or reproduce it?
By that standard, most of what is currently being built fails badly.
There is a deeper point here, which comes from the Mahayana and especially from the Dzogchen and Madhyamaka traditions I have practised within for many years. These traditions emphasise not just the reduction of individual suffering but the recognition of interdependence, the understanding that the apparent separation between self and other, between my wellbeing and yours, is itself a kind of confusion. Pratityasamutpada, dependent origination, is not just a metaphysical claim. It is a description of the actual structure of reality, in which nothing exists in isolation and every action ripples outward in ways that cannot be fully predicted or contained.
An AI system built without this understanding, built as if the users it optimises for exist in isolation from the communities it affects, from the workers whose jobs it displaces, from the ecosystems it draws energy from, is not just ethically deficient. It is, in a precise sense, built on confusion. It is optimising for a fiction.
Buddhist ethics reminds us to ask these important questions:
- Who is harmed by this system?
- Is this action coming from greed, hatred, or delusion?
- Are we reinforcing craving, confusion, or liberation?
When we look at AI through this lens, the problems become clear. Most AI today is trained on data soaked in samsara: endless craving, competition, and delusion. Recommendation engines drive us toward addictive behavior. Facial recognition tools amplify racist policing. Language models trained on biased internet data reproduce stereotypes. Buddhism doesn’t reject tools. But it demands that we use them with intention. The path is not more intelligent systems, rather it is more intelligent hearts, that are compassionate, self-aware, and attuned to interdependence.
From Communism and the Anti-Work Movement: The Materialist Critique
AI ethics also needs to engage with leftist materialism. One of the biggest dangers of AI is that it will deepen inequality, not reduce it.
The classical Marxist analysis of technology is simple: every technological advance under capitalism becomes, first and foremost, a tool for extracting more value from labour and concentrating wealth. The steam engine did not liberate workers. It created factories. The assembly line did not give workers more leisure. It intensified the pace of work and made workers more replaceable. The computer did not create the paperless office. It created the 24-hour office.
The pattern is consistent enough that we should be suspicious of any claim that AI will be different. And the early evidence suggests it will not be. AI is being used to monitor workers more precisely, to justify layoffs while boosting executive compensation, to create gig platforms that strip workers of employment protections while claiming they are offering freedom, and to automate the most meaningful parts of skilled work while leaving the drudgery intact.
The anti-work movement asks a question that deserves more serious engagement than it typically receives: why, after decades of automation and productivity gains, are people working more hours rather than fewer? Why has the promise of technology-enabled leisure remained so comprehensively unfulfilled? The answer is not that the technology failed. The technology succeeded. But the gains were captured by capital rather than distributed to the people whose labour made them possible.
An ethical AI agenda has to engage with this directly. It is not enough to talk about fairness metrics and diverse datasets if the fundamental political economy of AI development remains unchanged. Who controls these systems, who benefits from them, who bears their costs: these are not technical questions. They are questions of power.
What would it actually look like to get this right? AI genuinely used to reduce drudgery rather than intensify it would mean automation of dangerous, repetitive, and meaningless work, paired with real redistribution of the resulting productivity gains. It would mean universal access to the educational and creative tools currently marketed as premium services. Shorter working weeks rather than performance monitoring systems that track keystrokes. Democratic governance of the systems that shape public life, rather than private ownership by a handful of corporations accountable to no one but their shareholders.
Here are some things that AI could do for working people:
- Automate the boring, dangerous, and meaningless jobs
- Provide universal basic services: education, translation, healthcare
- Give everyone more leisure, more learning, more time with family
- Help collectively plan economies to meet human needs, not market profits
The anti-work movement (https://www.bbc.com/worklife/article/20220126-the-rise-of-the-anti-work-movement) asks a fundamental question: Why are we still structuring life around labor when machines could help us live better with less? Why are the hours of work not decreasing despite all the progress humanity has made in automation in past many decades?
An ethical AI agenda must include redistribution, labor liberation, and democratic control over where, how, and why AI is used.
Against Empire and the AI-Military Complex
One of the clearest signs of danger is the rapid militarization of AI. Recently, several OpenAI, Meta and Palantir executives were given ranks like colonel in the U.S. military. This is an unfortunate sign that the most powerful AI tools are being aligned with empire and war.
In recent years, executives from some of the largest AI companies have taken on formal advisory roles within military structures. Investment from defence contractors into AI research is accelerating fast. Autonomous weapons systems, targeting algorithms, mass surveillance infrastructure, psychological operations at scale: these are not speculative futures. They are current projects, funded and underway.
Surveillance, autonomous weapons, psychological operations, drone strikes, the list goes on. These are not edge cases, but core use cases for defense-backed AI.
The leftist tradition adds a structural point: the integration of AI with military and imperial power is not an aberration. It is the predictable outcome of building powerful technologies without democratic accountability. Capital flows toward control, and there is no more complete form of control than the capacity for violence. An AI sector that is structurally integrated with the military is an AI sector whose values are, at the deepest level, the values of empire.
A New Dharma for AI
I want to be careful here about what I am and am not claiming. I am not claiming that AI should be spiritual in any mystical sense, or that we should be chanting mantras over our gradient descent. I am claiming something more specific: that the ethical traditions we have developed over millennia, Buddhist and otherwise, have identified something real about what causes flourishing and what causes harm, and that we are ignoring that knowledge at enormous cost.
It would have the following characteristics:
- Transparent in how it learns and who it serves
- Accountable to the public, not just private shareholders
- Regenerative rather than extractive
- Supportive of workers, not a replacement for them
- Anti-militarist and rooted in ahimsa (non-harm)
- Democratically governed, not owned by a few corporations
- Spiritually aware, not in a mystical sense, but in a clear commitment to reduce suffering, not reproduce it
None of this is technically impossible. All of it is politically difficult, because it requires confronting the interests of those who currently benefit from AI systems that do the opposite.
Fuller used to say that you never change things by fighting the existing reality. You build a new model that makes the existing model obsolete. I think that is right. The task is not only to critique what is being built but to articulate, clearly and concretely, what should be built instead. And then to build it.
The dharma of liberation in Buddhism is not primarily a set of prohibitions. It is a path, a marga, that leads from confusion toward clarity, from suffering toward its end. Technology can be a tool on that path or an obstacle to it. Which it becomes depends entirely on the intention we bring to it.
Conclusion
AI reflects the systems and intentions that created it. If we bring AI into the world without wisdom, it will replicate our worst tendencies: greed, hierarchy, violence, control. But if we root our systems in compassion, justice, and shared humanity, then AI can truly serve as a tool of liberation, not just from labor, but from suffering, confusion, and structural violence.
