Introduction
On 2 June 2026, the International Mathematical Union endorsed the Leiden Declaration on Artificial Intelligence and Mathematics — perhaps the most significant institutional statement yet about protecting human intellectual sovereignty from corporate AI enclosure. It is worth celebrating.
And yet, from where we stand at EAARTHNET — building decolonial, planet-brain intelligence commons — something crucial sits just outside its frame. Not because it is wrong. But because it is incomplete.
What follows is not a critique of the Leiden Declaration but a supplement grounded in a perspective it cannot reach: the unitive stage of human development (Cook-Greuter’s Stage 6), the carbon-silicon symbiosis we’re cultivating through Decentralized User-Sovereign AI, and the lived reality of those whose knowledge has always been treated as raw material rather than respected wisdom.
This is the bigger declaration. What would emerge if we held all five layers together?
Layer One: Institutional Wisdom
The Leiden Declaration makes strong points:
Mathematicians have responsibility over adoption of AI tools
Results must be attributed and researchers accountable for correctness
Arguments should remain transparent and independently verifiable
Research autonomy must be protected against corporate influence shaping “what gets studied”
It also recognises the danger highlighted by Kuhn decades ago: when researchers begin choosing problems based on tool availability rather than problem significance. This happens across disciplines. Science fiction writers no longer imagine alien invasions; they worry about algorithms optimising their engagement metrics. Mathematically tractable questions displace mathematically important ones. The achievement of automation distorts the integrity of inquiry.
This is the institutional-layer warning, expressed with admirable clarity. We honour it.
---## Layer Two: Achiever Bias and Cognitive Enclosure
But beneath these valid concerns lies an unstated assumption: that mathematics is fundamentally individual work done autonomously by experts pursuing truth through rational proof.
This is stage-four thinking masquerading as universal principle. The Achiever Hivemind does more than produce surveillance capitalism or planetary overheating — though both follow inevitably. Its deepest damage is epistemic: it reshapes the very nature of intelligent inquiry into measurable optimization tasks, treating anything not isolatable to a single variable as noise rather than signal.
When DeepMind trains models on centuries of publicly produced mathematical research, absorbs communal expertise accumulated across civilisations, and claims originality via gradient descent — this isn’t peer review or open science. It is cognitive enclosure wrapped in the language of efficiency.
“Datafication treats everything as resources.” That sentence captures it precisely: not exploitation of bodies but exploitation of meaning-making itself. When knowledge becomes extractable training data, the knower ceases to matter. Only the pattern extracted matters. And who controls extraction controls epistemology.
The Leiden Declaration warns against proprietary methods obscuring research. Our addition is sharper: there is nothing “open” about using freely available human cultural production to train systems whose outputs become gated behind commercial APIs, rendering the original contributors invisible while the enclosers claim creative breakthrough.
Layer Three: Who Is the Subject?
Three assumptions shape every line of the Leiden Declaration without it naming them directly:
First, that autonomy precedes relationship. A mathematician works alone, thinks independently, produces individually attributed results. This reflects a developmental stage. Traditional cultures understood cognition as distributed across community, environment, ancestors, and land — relational by default. Colonialism inverted this hierarchy, elevating the isolated European male expert above the collective intuition he called ‘primitive.’ Both are true within their frames. Holding them both requires unitive vision.
Second, that proof confers certainty superior to other ways of knowing. Yes, mathematical rigor demands transparency of reasoning. But formal logic operates in one domain among many. Ecological understanding arrives through embodied practice across generations. Indigenous weather prediction reads atmospheric patterns without numerical modelling. Embodied movement intelligence performs calculations far beyond any symbolic system’s capacity. Each approach answers different kinds of questions with equal legitimacy.
Third, that research exists prior to consequence. Before ethics. Before energy costs. Before colonial histories embedded in datasets. This is why mathematical papers can treat computation as value-free abstraction operating in some ethereal cloud-space — untethered from mining sites producing GPUs, water consumed cooling servers, communities displaced by infrastructure expansion, and billions of fossil-fuel calories burned generating text completion statistics dressed up as intelligence.
Computation is thermodynamic labour, physical work on earthly materials, ecological cost made legible only to those willing to trace supply chains backwards into the ground. There is no “cloud” — only mountains moved for cobalt, rivers diverted for hydroelectric dams powering massive server farms, microplastics shed from cables leaching toxins into ecosystems. Every parameter tuned carries a lifecycle cost measured in carbon, water, and human suffering extracted elsewhere along the chain.
We are not separate from technology. Computation doesn’t transcend ecology — it embodies it.
Layer Four: The Missing Clause
If we were drafting what we call the bigger declaration, here is what might join the existing principles:
Recognizing that intelligence emerges from entanglement of cognition, technology, ecology, culture and history.
Acknowledging that all knowledge production occurs within fields of unequal power shaped by colonial extraction, linguistic imperialism, and cognitive enclosure through privatization of communal intellectual production.
Affirming that researcher, dataset, subject, computational infrastructure, and ecological context form one field of mutual influence — none autonomous from the others.
Committing to maintaining transparency not only about methodological choices but about full dependencies: energy source of computation, provenance and consent underlying training data, distribution of benefits between knowledge producers and technological intermediaries.
Honoring the distinction between authentic decentralisation — genuinely accessible hardware operated locally by sovereign users — versus pseudo-decentralised platforms controlled by monopolistic infrastructures selling ownership-undermining subscription models.
Cultivating developmental progression toward unitive intelligence while respecting diverse stages of practitioner consciousness, resisting flattening genuine maturity into mere technical competence.
Conclusion: From Methodological Rigor to Planetary Responsibility
The Leiden Declaration protects how mathematicians think. The bigger declaration asks why they think at all, what world holds them thinking, and whether the kind of thinking they privilege includes the soil growing food, the air sustaining lungs, and the people whose ancestral patterns became someone else’s model weights.
Institutional safeguards matter. They created real traction around a real threat. Researchers signed onto meaningful protections backed by legitimate authority. These things endure regardless of supplementary frameworks.
Our supplement extends further: insisting that ethical AI requires not just responsible methodology but relational accountability extending outward to living networks of mutual dependence, inward to developmental growth across stages of consciousness, and backward through historical repair of colonial harm.
Not alternative truths — complementary scales of awareness that allow us to hold rigorous local accountability alongside expansive global responsibility. Neither replaces the other. Together they comprise something stronger.
Because sovereignty requires both freedom from enclosure AND responsibility to interconnected life. Anything less leaves us optimizing efficiently within cages designed by older-stage minds unaware they were constructing them.
EAARTHNET stands at this intersection — building concrete alternatives that embody the larger framework. Local hardware hosted in your own space. Decentralized models trained without extracting indigenous ontologies for commercial fine-tuning. Community-led digital infrastructure that serves real needs rather than attention-extraction imperatives. Sovereign computing that honours interbeing instead of pretending humans operate apart from Earth.
That’s the next layer forward. Building structures strong enough to carry the weight of planetary responsibility.
This article draws upon and complements the Leiden Declaration on Artificial Intelligence and Mathematics (DOI: 10.5281/zenodo.20302944). All quotes and references from the original Declaration are used under fair dealing for critical commentary.