CosyAI seeks to build a true alternative to the corporate power game around AI, based on the principles of sincere cooperation. For this, we connect small and medium enterprises, cooperatives, open-source developers, and anyone else who shares the common purpose of responsible, human-aligned AI development, into a specifically designed cooperative framework.
Together, we can build a vehicle for coordinating efforts, pooling resources, and protecting ourselves from the negative effects of the “AI arms race”.
In this post I discuss the organizational framework for implementing this approach. It starts out from a basic question:
How can we create a structure that preserves our purpose, even if it grows big and powerful itself?
To achieve this, we need more than noble mission statements. All of the big power players claim to “benefit humanity” and their CEOs are touring podcasts and conferences to paint colorful visions of a future where “our problems will be solved”.
We also need more than the organizational provisions that OpenAI, Anthropic etc. use to claim that they are “aligned” with humanity’s fortune, while being owned by the same old corporate players and venture capital funds.
The organizational model we propose as an answer is the “Coop 3.0”. It’s designed to be a platform where a broad diversity of partners can each preserve their brand identity, business ambitions, and specific purpose, while coordinating resource pooling and strategic action at global scale.
No need to invent this from scratch
We build this model based on three powerful, practice-proven bodies of thought and practice, each addressing different layers of what makes traditional corporate companies go wrong:
Cooperative ownership solves the problem of extractive incentives. The cooperative model is over 150 years old, but its core principles remain as relevant as ever.
Anti-corruption design addresses the problem of power concentration and silent purpose erosion. We reuse the principles of anti-corruption for designing the original Coop 2.0 model, which includes mechanisms for ensuring transparency, accountability to all stakeholders, and a culture of integrity among all people in the organization.
Self-organizing principles address the problem of dehumanizing complexity that comes with growing size and scope of organizations and allow for a new level of alignment with our deeper human traits. For this, we heavily rely on the concept of “Teal organizations”, based on the research by Frederic Laloux. In his book Reinventing Organizations, he shows that companies can operate effectively through self-managing teams and practices that let people show up as whole human beings.
A pragmatic approach
Together, these three steps mark the evolution to the “Coop 3.0”:
The traditional cooperative model established democratic ownership.
Coop 2.0 adds corruption-resilient design.
Coop 3.0 adds the human dimension: the organizational conditions for people to cooperate sincerely.
To complement these layers, we picked elements from DAOs (Decentralized Autonomous Organizations, which offer tools for transparent governance and resource allocation at scale) and from the more advanced version of DisCOs (Distributed Cooperative Organizations, which bring Open Value Accounting, feminist economics, and commons-oriented coordination).
We tried to do this as ideology-free as possible. We are not committed to any specific technology or any specific political view. What counts is the outcome of an organization that allows humans to thrive and achieve the purpose they really aim for. At the same time, a Coop 3.0 is a learning organization that adopts new concepts and practices along the way, always under the premise of preserving the organization’s real purpose.
Better economic and social outcomes
It’s important to understand that this is not about a non-profit approach or about overcomplicating business. In his research, Frederic Laloux has shown that Teal organizations consistently perform better.
Classic corporations and the traditional “Silicon Valley model” of startups are optimized for rapid growth of the speculative valuation of a company, not for the actual value created for society.
Laloux identified the specific practices and organizational designs that make purpose-driven self-organized companies superior to classical corporations, both as workplaces and in terms of achieving their purpose.
He also identified two make-or-break conditions for their survival: the CEO must hold the space for the model, and the board must understand and embrace it. If either fails, the organization faces a great risk of regressing to old command-and-control the moment a crisis hits.
This seems to be a dilemma of self-organizing companies, making them especially dependent on the right people being in charge.
The Coop 3.0 model addresses this by baking the conditions for self-organization into the ownership and governance structure itself. Cooperative ownership means no external investors can panic and override the model. Anti-corruption mechanisms prevent internal power concentration and (self-)deceptive dynamics.
The seven pillars of Coop 3.0
Pillar 1. Cooperative ownership
The cooperative model solves, by design, two of the most destructive features of traditional corporations: the concentration of power in the hands of a few and the compulsion to grow endlessly. The seven cooperative principles have proven valuable and practical for over 150 years and millions of cooperative businesses worldwide.
For CosyAI, this is the foundation that makes everything else possible. Data pooling becomes possible, even for sensitive data, because the members control the very organization that manages it. Infrastructure investments can be shared because no single investor can capture the returns. Instead of every member company buying and setting up their own local server infrastructure, CosyAI will bulk purchase GPUs and other hardware and offer a one-click setup at much lower cost.
Partner organizations can preserve their own identity and business model while participating in collective strategy, because the governance structure guarantees that no one can accumulate disproportionate power.
Cooperative governance and accountability:
In a cooperative, every member has one vote in the general assembly, regardless of how much they invest. Nobody can buy control. Cooperatives are accountability models rather than management models, so daily operations, product development, and business strategy can be handled by specialized teams with full autonomy. The difference to a classical startup is that management is accountable to a broad base of members, rather than to a handful of majority investors chasing returns.
In a Coop 3.0, members elect a supervisory board that oversees the management of the cooperative and can suspend the managing board if it doesn’t comply with the provisions of the bylaws. But the members of the managing board, including the CEO, are elected by the employee assembly, thus making sure that the coop is run in alignment with the collective wisdom and insights of those who do the actual daily work. This way, the managing board is also bound to help establish a healthy and empowering work culture.
A Coop 3.0 is a multi-stakeholder cooperative: membership is open to anyone who feels affected by the company’s activities. Workers, users, partner organizations, neighboring communities. Shares are kept affordable to keep the threshold low. The company can grow its membership for as long as it makes sense, but it is never desperate for growth.
Protection from chronic growth compulsion:
This is ensured by a crucial legal feature that differentiates cooperatives from classic corporations: the price of a share is fixed in the bylaws. This single provision eliminates the entire machinery of financial speculation and the pressure to inflate shareholder value at any cost. When a cooperative company is economically successful, more members join and buy shares at the fixed price. If it slows down, members can decide to cancel their shares and get their investment back.
Since the only return on investment is the dividend, cooperatives are actually real “for-profit” companies. If the company grows, it means that it is actually providing value at efficient costs. No one has an incentive to push for growth purely for the sake of growth. If the dividend increases compared to the risk, new members have a reason to join. Once the market is saturated, this process ends, and the company can happily continue at its current scale.
If all stakeholders are represented in the membership, no one has an incentive to cut costs by outsourcing them on one stakeholder group.
Pillar 2. Corruption-resilient design
Even the best ownership model can be undermined if the organization lacks safeguards against the accumulation and impact of power. Three basic principles of anti-corruption, originally developed for fighting corruption in governments and public institutions, can be applied to any organization: transparency, accountability, and integrity.
Transparency means that any information that can be published legally and without harming privacy must be published. This includes finances, decision-making processes, and the rationale behind strategic choices. When everything is visible, it becomes much harder to quietly serve private interests.
Accountability means that everyone in the organization can be held accountable if they deviate from the common purpose. Management is elected by employees, allowing the people working for the organization to make sure that all activities are aligned with its purpose. Clear basic principles are protected by the bylaws and can only be changed with a 90% majority.
Integrity means designing the organization so that it doesn’t incentivize greed in people. Legal provisions include defined maximum salary ratios, caps on individual shareholdings, and a fair, pre-defined distribution of surplus. If there is nothing disproportionate to gain, the incentive to corrupt is removed. Laloux’s research points the same way: Teal organizations cap the ratio between highest and lowest pay and mostly avoid individual bonuses, since extrinsic incentives distract from inner motivation and breed negative competition.
Pillar 3. Self-organizing teams
For CosyAI, this is what makes this model especially usable as an AI lab with various research, infrastructure, and product streams. The self-organization model allows dozens of autonomous teams working on different focus topics, at different paces, with different expertise, while still serving a common purpose. Some teams concentrate on the underlying GPU infrastructure. Other teams do foundational research on “cosy” AI models. Still others develop privacy-preserving AI tools or design the user experience for data-sovereign personal AI agents.
These teams need real autonomy for the kind of deep work they do. At the same time, the organization allows them to share process innovation, tools, resources, and challenges that outgrow a single team’s capacities. Teams draw their own boundaries, while inter-team coaches facilitate cooperation and exchange.
This is part of the Teal model of organizations that is characterized by three breakthroughs:
Self-management. Anyone in the organization can make any decision, provided they first seek advice from all affected parties and relevant experts. They don’t need consensus, but they must genuinely listen. Conflict is resolved through structured peer processes: first a private conversation, then mediation by a trusted colleague, then a panel if needed.
Wholeness. People can drop the professional mask and bring their full humanity to work. Meetings start with check-in rounds and employees get a range of supportive practices to connect with their individual purpose. The chronic stress of performing a role gradually dissolves.
Evolutionary purpose. Instead of top-down planning and sales targets, the organization operates through a “sense and respond” model, and the purpose emerges from the collective rather than being imposed by leaders. Buurtzorg, a Dutch home-care organization with more than 10,000 employees studied by Laloux, openly shares its methods with competitors, on the insight that anyone helping to achieve the shared purpose is an ally rather than a threat.
One lesson from decades of cooperative and activist experience: declaring an organization “flat” is a recipe for invisible power games. Self-organization requires deliberate structures, clear agreements, and explicit processes. Without them, power concentrates informally and burns out the people who carry the most weight.
Pillar 4. Values-based coordination
In most organizations, only the work that generates revenue counts. Everything else, the mentoring, the emotional labor, the community building, the unglamorous administrative tasks that keep the whole operation running, remains invisible. The people who care the most carry the heaviest unseen load, while the system rewards those who optimize for measurable output.
For CosyAI, this takes on a global dimension. Some partners are established European SMEs. Others are open-source developers in Nairobi contributing code in their free time, or community organizers in São Paulo running local AI education workshops. If we only recognize and reward the work that directly generates revenue, we replicate exactly the extractive logic we set out to replace.
The DisCO model offers a framework for this. Grounded in feminist economics, DisCO developed an Open Value Accounting system that makes visible what traditional organizations ignore. It tracks three distinct value streams:
Livelihood work (the goods and services that generate revenue),
Love work (voluntary contributions to the commons, like open-source code, educational resources, and shared knowledge), and
Care work (the affective and administrative labor that maintains the collective itself, from facilitating meetings to supporting a struggling colleague to handling the thankless tasks everyone relies on).
Revenue is pooled and distributed dynamically, and members who take on more care work receive compensation adjustments from those who take on less. For CosyAI, this framework will feed into the cooperative basic income pool: a share of collective revenue distributed to all human members, recognizing that the value of the cooperative is created by the entire community.
Pillar 5. Organic scaling into a federated network
Growth doesn’t need to mean scaling up a single entity. Following the DisCO principle of federation, Coop 3.0 organizations can expand by replicating: if practically useful, a separate cooperative can be created for a new product line or research focus. They can start with small nodes of around 15 to 20 people that mutualize economic power through lightweight shared protocols. These nodes connect in an overarching cooperative framework that gives them collective bargaining power and allows for coordinated action at eye level with global corporations.
Similarly, CosyAI encourages the creation of local AI coops that facilitate local data ownership or run community data centers.
Pillar 6. Sincere cooperation
A Coop 3.0 is a Cosy Company: an organization explicitly designed to support the shift from cooperating instrumentally to cooperating sincerely.
The starting point is a biological fact: our emotional architecture evolved for cooperation. Trust, compassion, and the capacity to care about each other are essential features of the human species. But the conditions of our current societies constantly push us into a latent fear state, leading to chronic defense-mode and competition thinking, and the feeling that we need to look after ourselves first.
This is one reason why work in cooperative organizations can often feel slow, inefficient, and exhausting. No perfectly designed organizational structure can resolve it, if the people inside remain trapped in the defensive patterns that our societies have imprinted on us. The Cosy practices therefore focus on how we relate to ourselves and to each other. A Cosy Company creates a space where people can gradually reconnect with their sincere core and discover that cooperating from a place of trust and love is not naive, but the most effective and deeply fulfilling way to work. When the pressures of performing, competing, and protecting yourself fall away, an enormous amount of energy becomes available for the actual work and for genuine human connection.
The Cosy layer builds on the “wholeness” practices documented by Laloux and adds practices for going deeper, at the level of the individual, the team, and the organization as a whole. Many people in our coalition work professionally in coaching, facilitation, and contemplative practice, and they can guide members and partner organizations in building and maintaining practices that fit their own context.
Pillar 7. Embracing AI as an organizational tool
CosyAI will use AI across every dimension of the organization. This makes it even more important to host the models, the agentic frameworks, and the data for specialization on our own, sovereign hardware infrastructure.
AI agents can be used for accountability and oversight tasks: monitoring governance compliance, tracking financial flows, flagging inconsistencies. This solves the “overview problem” in cooperatives. Members simply don’t have the time and capacities to oversee every decision and activity. But they can define their base values and priorities in a natural language conversation with their personal “participation and accountability agent”. This can also help to prevent governance from becoming political and dominated by the loudest voices with the most free time.
One rule is non-negotiable: for every AI agent deployment, the AI also creates an educational session for team members to understand the underlying process and logic behind it. If the AI develops more complex processes, it must ensure that our human understanding grows with it. Major insights are released to the public under appropriate open-source licenses. In this sense, a Coop 3.0 is always also a research organization.
At the same time, a Coop 3.0 explicitly protects against the risks of AI: cognitive degeneration through over-reliance, the dehumanizing effects of replacing human connection with machine interaction, and the psychological harms that can come from treating AI systems as companions or emotional substitutes. The Cosy layer keeps people grounded in human connection while using AI as a powerful tool.
Rewarding early contributions
The old corporate model does have an advantage: it’s an effective vehicle for collecting upfront investment before the company actually becomes profitable, which matters in investment-heavy fields like AI. In reality, this risks turning the company into a capital accumulation machine where the game is increasingly about externalizing costs to society and avoiding competition by chasing “unfair advantages”.
For CosyAI, we need to find a way to raise enough resources to face big tech at eye level, while preserving the cooperative principles. We do so by using a transparent system that rewards early commitment. It compensates members who invest time and money early and carry more risk, even when dividends are only expected further down the road.
The goal is to make the process clear, verifiable, and easy for anyone to check so they can make grounded decisions about whether and when to invest. This mechanism is called Fair Founding Points, or FFP.
The Fair Founding Point mechanism
Important: Note that the following is a first proposal and comments and suggestions are invited (just use the reply function below the post). The final mechanism will be reviewed and decided before we actually register the cooperative.
FFP are non-transferable internal accounting units, written into the statutes. They carry no voting weight and they cannot be sold, traded, or priced by a market, so they create no speculative equity and no secondary market.
In any year in which the cooperative generates a surplus, allocation follows a fixed protocol. Of the distributable surplus, 25% is distributed proportionally across all regular member capital shares, and 25% goes to the FFP pool, distributed to holders in proportion to their FFP balance. The remaining 50% will be reinvested and used for supporting the creation of local nodes.
FFP for member shares:
The FFP are used to reward the people who carry the risk and the weight of the founding work. For every euro invested in cooperative shares in each founding phase, a member gets an additional number of FFP as outlined in the overview below.
FFP for active contributions:
Similarly, for every hour of time contributed, a member gets a certain amount of FFP allocated through a mechanism that is similar to classic employee stock options. Accredited developers, researchers, organizers, and other helpers earn FFP for defined work packages reviewed and approved by their co-workers.
Issue rates based on founding phases:
To reward share investments in the early startup phases, FFP rates are higher in the beginning and decline with the increasing establishment of the cooperative. The phases and rates are fixed in the bylaws. The phases also reflect the different focus groups that are actively addressed in our outreach to join the coop.
There is also a maximum investment per member in each phase. If the cap for a phase is reached, remaining investments will automatically be accounted for the next phase. This allows the coop to collect substantial funding by bigger organizations, while fully appreciating the contributions of individual members and smaller organizations at the earlier stages.
FFP collected through active contributions contribute to a member’s account for each phase and are also subject to the cap.
The FFP Phases
Phase 1: First €100k raised
Rate: 20 FFP per €1 invested.
Max Investment: €5,000 per member (100k FFP).
Active contributions: 400 FFP per hour.
Focus Group: Value-aligned early supporters, individual contributors, small cooperatives, SMEs, and local community spaces.
Phase 2: Next €900k raised
Rate: 10 FFP per €1 invested.
Max Investment: €50,000 per member (500k FFP).
Active contributions: 200 FFP per hour.
Focus Group: Established SMEs, successful local cooperatives, regional NGOs, and angel syndicates looking to secure early sovereign infrastructure access.
Phase 3: Next €4M raised
Rate: 5 FFP per €1 invested.
Max Investment: €500,000 per member (2.5m FFP).
Active contributions: 100 FFP per hour.
Focus Group: Municipal funds, mid-sized enterprise consortiums, and local energy cooperatives.
Phase 4: Next €95M raised
Rate: 3 FFP per €1 invested.
Max Investment: €5,000,000 per member (15m FFP).
Active contributions: 60 FFP per hour.
Focus Group: National cooperative banks, state-level infrastructure funds, EU matching grants, and deep-tech syndicates.
Phase 5: Next €900M raised
Rate: 1 FFP per €1 invested.
Max Investment: €25,000,000 per member (25m FFP).
Active contributions: 20 FFP per hour.
Focus Group: Sovereign wealth funds, international ESG pension funds, massive EU structural initiatives, and mass-market SME adoption globally.
Starting the cooperative basic income pool
The five phases add up to one billion euros of financial investments through member shares. In this sense, our fundraising aims at becoming the first “billion-euro crowd-investing campaign”.
Once the target of one billion euros is reached, the 50% pool of surplus allocation for reinvestment and network building is split. 25% stays with reinvestment and network building, the other 25% starts the cooperative basic income pool. All nodes of the network will equally contribute 25% of their surplus into this pool, which will be used to give all human members in the network a basic income to help free them from the economic pressures of the old system.
Guardrails and sunset clause.
Total issuance is capped at two billion FFP. The whole mechanism carries a strict ten-year sunset clause, starting from the moment that the cap is reached. When it expires, all outstanding FFP expire with it, and the 25% surplus allocation is transferred permanently tothe basic income pool. The legal characterization of FFP is still being verified and has to be settled before registration.
Current phase
CosyAI is currently forming a coalition around this shared vision and preparing for the registration of a European Cooperative Society (SCE), a legal form that allows for global membership under EU data protection standards.
So far, more than 800 people, organizations, and companies have signed up.
This is the stage where the basic setup gets shaped. If you want to help build a real alternative to the corporate power games, this is the moment to join.
We are all needed
If this resonates with you, please reach out. We’ve been trying to put together concrete options for contributing at different levels:
If you are responsible for an organization, you can join the CosyAI coalition as a founding member.
As an individual, you can become a contributor member and support the creation of CosyAI and other Coop 3.0 pilots.
We don’t need to convince everyone at once. But we do need everyone who clearly feels that this is the right thing to do now.