WSIS Forum 2026
AI-generated report
ENFR

Governing the Systemic Impacts of AI Where Regulation Doesn't Reach: Standardised Commitments and Lateral Requirements across Value Chains

8 speakers
Summary

The discussion centres on the AI Commons (AIC) Framework, a voluntary, bottom-up governance instrument designed to facilitate beneficial systemic impacts across AI value chains, developed by a working group housed at the Free University of Brussels . Dr. Em Lenartowicz opened by critiquing existing AI governance approaches, noting that the United Nations scientific panel has counted over 40 governance instruments, most of which are fragmented, corporate-focused, and rarely measure real-world effectiveness . She argued that corporate-level governance risks replicating the same extractive and exclusionary dynamics already seen in human-led organisations, and that the AI transition demands a higher standard .

The AIC Framework offers a compact, human-readable profile through which AI actors can voluntarily commit to six dimensions: Reciprocity, Sustainability, Openness, Governance, Access, and Value Sharing . Crucially, commitments can be extended upstream and downstream along value chains, creating mutual expectations and competitive pressure towards higher standards, analogous to the self-organising dynamics seen in Creative Commons and open-source licensing . Dr. Ann Borda highlighted the framework's potential as an operationalisation of responsible AI principles and drew a parallel to the concept of a "social licence to operate," noting that licensing offers a lightweight pathway within innovation cycles already supported by bodies such as the OECD .

Dr. Mihaela Ulieru presented SingularityNet as a real-world example, illustrating how its open-source, decentralised platform embodies several AIC dimensions, including distributed computing for sustainability and blockchain-enabled value sharing . Dr. Yang Yoon emphasised the need to incentivise individual commoners and small-to-medium businesses, proposing cooperative structures and distributed learning algorithms as pragmatic entry points .

An audience question raised the issue of safety and liability, to which Dr. Lenartowicz responded that the framework complements rather than replaces existing safety regulation, focusing instead on systemic effects that fall through current governance gaps . On enforceability, she explained that the framework relies on ordinary contract law, with commitments embedded in standard legal documents such as terms of service and procurement contracts, removing the need for new centralised institutions . The discussion concluded with broad agreement that voluntary, decentralised frameworks grounded in existing legal infrastructure represent a promising and pragmatic pathway towards more equitable and accountable AI deployment.

Keypoints

  • Overall Purpose

  • The discussion centres on the presentation and exploration of the AI Commons Framework (AIC Framework), a voluntary, bottom-up governance tool designed to facilitate self-organisation across AI value chains. The goal is to move beyond existing corporate-level AI governance instruments and instead induce beneficial systemic effects by encouraging AI actors to make and extend commitments around reciprocity, sustainability, openness, governance, access, and value sharing.
  • --

  • Major Discussion Points

  • Limitations of existing AI governance frameworks and the need for a different approach: Dr. Lenartowicz argues that over 40 existing AI governance instruments are fragmented, corporate-focused, and rarely measure real-world effectiveness. The AIC Framework is positioned as a complement to these, targeting systemic effects that fall through the cracks of conventional regulation - particularly the externalities and inequalities that emerge from the aggregated behaviour of many AI actors, even when each individually complies with the law.
  • The six commitment pillars of the AIC Framework (RSOGAV): The framework offers AI actors a voluntary set of commitments - Reciprocity, Sustainability, Openness, Governance, Access, and Value Sharing - which can be applied to their own operations and extended upstream and downstream across value chains. Each commitment is represented visually in a compact, legible icon, similar in concept to Creative Commons or open-source licensing. Dr. Ulieru illustrated how SingularityNet already embodies many of these pillars through its decentralised, open-source AGI platform, including distributed computing for sustainability and blockchain-based revenue sharing for reciprocity. - The value chain multiplication effect and bottom-up enforcement: A central design principle is that individual commitments alone are insufficient; the framework is intended to create exponential positive dynamics by enabling AI actors to require the same standards from their upstream suppliers and downstream operators. Enforcement relies not on new institutions but on existing contract law - commitments embedded in terms of service, API conditions, and procurement documents become legally binding obligations that stakeholders can act upon.
  • Inclusion of small and medium-sized enterprises (SMEs) and individual commoners: Dr. Yoon raised the challenge that individuals and SMBs are largely excluded from the AI economy, dominated by big tech and a small number of technocrats. He proposed cooperative models using distributed learning algorithms to allow shared data contribution without exposing sensitive information, with profit sharing once AI services are created - suggesting a pragmatic, phased approach starting with SMBs before extending to individuals. - Safety, accountability, and the role of openness: An audience member raised the question of liability and standardised safety assurance across jurisdictions. Dr. Lenartowicz acknowledged that safety is largely addressed by existing regulation, with the AIC Framework focusing on systemic effects beyond compliance. Dr. Ulieru argued that openness itself constitutes a safety mechanism, as transparent, decentralised, and collectively scrutinisable systems are inherently more defensible than opaque, closed AI systems.
  • --

  • Overall Tone

  • The overall tone of the discussion is collaborative, intellectually engaged, and cautiously optimistic. Dr. Lenartowicz sets a constructive and reform-minded tone from the outset, framing the framework as a pragmatic response to well-documented failures in existing AI governance. Panellists contribute with genuine enthusiasm and real-world examples, maintaining a collegial atmosphere throughout. There is a moment of productive tension when Dr. Ulieru expresses a philosophical disagreement with the Universal Basic Income component of the Value Sharing pillar, preferring a meritocratic model - though this is handled respectfully and used by Dr. Lenartowicz to illustrate the voluntary nature of the framework. Audience questions introduce a slightly more sceptical note, particularly around enforcement and safety accountability, but these are addressed constructively. The tone remains consistently forward-looking and solution-oriented throughout.
Speakers Overview
DE
Dr. Em Lenartowicz
124 wpm · 26 min
AG
Ana Garcia Robles
140 wpm · 3 min
DA
Dr. Ann Borda
128 wpm · 4 min
DC
Dr. Cristian Axenie
142 wpm · 2 min
MU
Mihaela Ulieru
144 wpm · 6 min
DY
Dr. Young Yoon
114 wpm · 3 min
A
Audience
149 wpm · 2 min
A2
Audience 2 - Sidoin Sotudo-Aiti
161 wpm · 40 s

Expanded Summary: AI Commons Framework - Session Presentation and Panel Discussion

#

Introduction and Context

The session was dedicated to the presentation of the AI Commons (AIC) Framework, a governance instrument designed to facilitate bottom-up self-organisation across AI value chains with the intention of producing beneficial systemic impacts from the AI transition . The framework is being developed by a working group housed at the Free University of Brussels, five members of which were present at the panel . Dr. Em Lenartowicz opened proceedings by noting that the session was also intended to gather feedback, expressions of interest, and networking, given that the framework remains under development . The working group overlaps partially with the steering committee of the AI for Good impact initiative .

#

Critique of Existing AI Governance Frameworks

Dr. Lenartowicz began by acknowledging that the phrase "AI governance framework" risks provoking a sense of fatigue, given the proliferation of such instruments . She cited the United Nations scientific panel's count of over 40 governance instruments, noting that a recently released report from that body found these frameworks to be mostly fragmented, concentrated at the corporate level, and rarely measuring real-world effectiveness . The AIC Framework, she argued, is fundamentally different in orientation: rather than measuring effects, its focus is on inducing positive systemic effects and ensuring they propagate .

A central critique underpinning the framework's design is that corporate-level AI governance carries an inherent limitation. Most existing instruments appear to aim for a level of controllability in AI-run business operations comparable to how human-led organisations behave . Dr. Lenartowicz argued that this standard is insufficient, since human-led organisations have themselves generated significant problems - including extraction, accumulation, and the creation of externalities and exclusion - and replicating these dynamics through AI will only magnify them through acceleration . The challenge and design task of the framework, therefore, is to enable new types of AI-involving organisations to operate at a higher standard than human organisations have historically achieved .

#

The Logic of Exponential Dynamics and Multiplication

A key conceptual contribution of Dr. Lenartowicz's opening was her distinction between additive and multiplicative dynamics. She argued that exponential dynamics - whether wanted or unwanted - do not emerge from the mere aggregation of similar projects, but from the multiplication of interactions among actors . In the business context, the tendency to extract, accumulate, and create externalities does not arise from the isolated strategic decisions of individual corporations, but from competitive dynamics and market pressures that make it practically impossible for any single actor to behave differently from others . This insight reframes governance not as a matter of individual compliance but as a matter of emergent systemic behaviour.

Crucially, Dr. Lenartowicz argued that if humanity is capable of generating unwanted exponential dynamics, it is equally capable of deliberately designing beneficial ones . The question becomes: how can a practice and culture be established in which AI actors hold each other to higher standards and create mutual pressure to operate on higher ground ? This logic directly motivates the framework's design, which is intended to harness the multiplication effect for beneficial ends rather than merely constraining harmful ones.

#

The Structure and Six Pillars of the AIC Framework

The AIC Framework is designed as a compact, human-readable object - described as something like a QR code but understandable at a glance - through which any AI actor can voluntarily express commitments across six dimensions . These dimensions are represented by the acronym RSOGAV: Reciprocity, Sustainability, Openness, Governance, Access, and Value Sharing . The framework is explicitly positioned as a bottom-up and voluntary instrument, analogous in design to Creative Commons or open-source licensing, where using an icon signifies acceptance of standardised legal terms .

The visual structure of the framework uses horizontal lines for the six commitment types and vertical positioning to indicate where in the value chain a commitment lies. A central dot represents a core commitment within the actor's own deployment, whilst positions to the left and right represent expectations from upstream suppliers and conditions placed on downstream operators respectively . This architecture enables actors not only to make individual commitments but to extend those commitments into mutual requirements across their value chains, creating the reciprocal pressures and mutual expectations that can generate beneficial systemic dynamics . As a concrete illustration of this value chain logic, Dr. Lenartowicz noted that the Artificial Superintelligence (ASI) Alliance is one entity to which the upstream-downstream value chain logic can be applied [S1].

Each of the six commitment types carries a specific and ambitious meaning. The Sustainability (S) commitment requires that the environmental footprint and cost of the entire operation within the boundaries of a deployment balances with its effects, at minimum in a neutral way and ideally in a regenerative way . Dr. Lenartowicz offered the example of solar-powered robots planting trees in deserts as an illustration of a deployment whose regenerative environmental effect can be measured against its operational cost . The Openness (O) commitment extends beyond open-source code to include the operational level of an AI deployment - making weights, internal AI procedures, and incident-handling processes open to civil society scrutiny . The Governance (G) commitment requires that the full representation of all stakeholder groups be present at the decision-making table for key aspects of a deployment. As an illustrative example, Dr. Lenartowicz cited the "sunset decision" - the decision to close a system when it is deemed harmful - as one that must not rest solely with owners but must be open to all stakeholders .

The Access (A) commitment involves making the full set of AI capabilities available free of charge to everyone, a standard that Dr. Lenartowicz suggested is particularly appropriate for publicly funded projects . The Value Sharing (V) commitment is framed as a contribution to a universal basic income scheme, targeted at regions with the highest concentration of people living in extreme poverty, with legal protections ensuring data is not shared inappropriately . The framework also includes a commitment type under which a project voluntarily chooses to share a fraction of its gross revenues with broad classes of contributors - creators, data producers, and knowledge workers - who have contributed to the system's capabilities; this is presented as an illustration of how the framework's commitments operate in practice rather than being definitively assigned to a single named pillar . The framework allows actors to choose which commitments to adopt and where to apply them across their value chains, making it a flexible rather than prescriptive instrument .

#

External Validation: Ana Garcia Robles and the UN Perspective

Ana Garcia Robles, participating via a pre-recorded message from a parallel session at Palexpo, brought the perspective of the United Nations Office for Digital and Emerging Technologies, where she leads a newly established lab on AI governance . She confirmed that recent UN work on AI governance and practices with the private sector had generated insights that resonate strongly with the AIC Framework. In particular, she noted the shift from governing more static AI systems to addressing the dynamic nature of modern AI - especially given the emergence of agentic technologies - as a challenge that the framework addresses well . She also affirmed the UN's finding that AI governance lives not only in norms and their implementation but also in decisions, interactions, value chains, and ecosystems, including procurement decisions made by customers and collaborators .

Garcia Robles further noted that the UN has been conducting significant work on governance interoperability, and that the six pillars of the AIC Framework resonate with pathways identified through that work . She expressed a desire to continue collaborating and to explore opportunities to test and pilot the framework, signalling institutional openness to engagement .

#

Operationalisation and Licensing: Dr. Ann Borda's Contribution

Dr. Ann Borda, introduced by Dr. Lenartowicz as the head of the newly established AI Governance for Humanity Lab of the United Nations , framed the AIC Framework as an operationalisation of responsible AI principles - many of which have been articulated in OECD recommendations and other UN agency guidance - that have so far remained largely aspirational . She highlighted the importance of the framework in the context of an uneven regulatory landscape and a push towards self-regulation, arguing that licensing offers a lightweight yet viable pathway that fits within innovation cycles already supported by governments and bodies such as the OECD .

Dr. Borda drew attention to digital public infrastructure (DPI) as a particularly promising area where the framework could be applied, noting its potential to channel public procurement towards responsible AI and to limit the ability of larger corporations to consolidate power through vertical integration . She discussed the spectrum of existing licensing models - from permissive open-source licences to more restrictive Creative Commons arrangements - and noted that these are already being shaped by the AI ecosystem, making it important to think carefully about how licensing arrangements can be improved . She raised a significant caveat about the risk of dual licensing, where a commitment profile could become a checklist for one purpose whilst monetisation occurs under a different, less responsible licence .

Perhaps most evocatively, Dr. Borda drew a parallel to the concept of a "social licence to operate," historically used in the mining industry, where communities and companies developed a dynamic relationship that granted or withheld legitimacy based on community acceptance . She suggested that the AIC Framework can be understood as a contemporary form of social licence to operate for AI deployments, grounding the abstract framework in a well-established historical precedent .

#

SME Adoption and the Pace of AI Growth: Dr. Cristian Axenie

Dr. Cristian Axenie, from the Nuremberg Institute of Technology, contributed a perspective focused on the practical adoption of the framework by small and medium-sized enterprises (SMEs). Note that his opening remarks were partially inaudible or cut off in the transcript. He observed that many SMEs are looking for ways to add value to their customers through AI, and that there are already examples of companies delivering not just a product or system but a certificate on the product - a development he sees as a step towards harmonising deployment perspectives . He argued that the AIC Framework contributes to this harmonisation effort .

Dr. Axenie emphasised the importance of awareness and acceptance as preconditions for adoption, noting that many SMEs are aware that AI can bring value but lack a clear recipe for how to adopt it responsibly . He also extended Dr. Lenartowicz's mathematical framing of exponential dynamics, arguing that governance measures must grow at least as fast as AI systems themselves in order to handle the pace of AI development . He presented the AIC Framework's pillars - openness, sustainability, access, value, and reciprocity - as providing both the tools and the means for pragmatic adoption, and expressed confidence that the framework is on a good path to building awareness .

#

SingularityNet as a Real-World Case Study: Dr. Mihaela Ulieru

Dr. Mihaela Ulieru, from SingularityNet, presented the platform as what she described as a "poster child" for the AIC Framework - a real-life project that already embodies many of its pillars in practice . SingularityNet is described as the only open-source, decentralised artificial general intelligence platform, accessible to everyone and comparable to "the Linux of AGI" . The platform's Hyperon system is fully open source, allowing anyone to see what it does and access it .

On the dimension of reciprocity and data sharing, Dr. Ulieru explained that the platform operates through what she described as the Artificial Super Intelligence chain (ASI chain): a decentralised architecture that enables data to remain private whilst being encrypted via blockchain technology and pooled with others' data to generate better results, with contributors being paid for the data involved in any given product . Similarly, developers who post algorithms on the platform are automatically compensated when others use them, via blockchain and encrypted technologies . On governance, SingularityNet operates through a constitutional collective based on reputation, where decision-making weight is commensurate with expertise, allowing stakeholders to delegate decisions to those with greater relevant knowledge . On sustainability, the platform's distributed computing architecture eliminates the need for large data centres, achieving sustainability through diffusion of computing power across the network .

However, Dr. Ulieru introduced a notable point of disagreement with the framework's Value Sharing pillar. She stated that SingularityNet sees value as meritocracy - reward commensurate with contribution - and organises hackathons to reward real work rather than endorsing a universal basic income model . She acknowledged having had "big discussions" with Dr. Lenartowicz on this point . She did, however, add a humanising qualification: reflecting on the pace of AGI development and the fact that AI systems are beginning to surpass human experts in some domains, she remarked that she may be reconsidering her position on UBI [S2]. Dr. Lenartowicz responded by welcoming this as a "wonderful illustration" of the framework's bottom-up and voluntary nature, emphasising that the framework allows actors to express what they believe in and lets their environment determine whether it is attracted to that choice of value . Dr. Ulieru also argued that openness itself constitutes a built-in safety mechanism, describing the platform as "the best immune system on offer today" because it is transparent, decentralised, interpretable, and collectively defended .

#

Inclusion of Small Players: Dr. Yang Yong

Dr. Yang Yong, from Hong Kong University, shifted the discussion towards the group he described as most important: individual commoners and small to mid-sized businesses left out of the AI economy, which is heavily dominated by big tech and a small number of technocrats . He argued that these small players will not participate unless incentivised with sizable rewards given upfront upon the contribution of their personal data, but that the time required to create value-added AI services from such data makes immediate reward practically very difficult . He noted the irony that small players already give full consent to big tech to share all their personal information whilst receiving no compensation, making their concern about privacy within a cooperative model somewhat paradoxical .

As a potential solution, Dr. Yong proposed the creation of cooperatives within the AI commons framework, where cohorts of shared business interest could gather together, pool private resources, and create shared value-added services . He suggested that distributed learning algorithms could be used to circulate AI models through different sources of private information without exposing sensitive data, thereby addressing privacy concerns . Once AI models are created through this process, profit sharing could follow . However, he acknowledged that even within this cooperative model, rewarding individuals upfront remains very difficult, and proposed a pragmatic, phased approach: begin with SMEs, learn the principles, and then extend participation to individuals .

#

Audience Questions: Safety, Accountability, and Enforcement

The first audience question raised the issue of safety and liability, asking who bears responsibility when an AI model causes harm and whether this should be incorporated into the licensing process . The questioner called for a standardised tool applicable across jurisdictions and industries - analogous to financial reporting standards or nutrition labels for food - to make safety and governance assessments simple and universally understandable .

Dr. Lenartowicz responded by clarifying that the AIC Framework's focus is on complementarity - addressing systemic effects that fall through the cracks of existing safety-focused governance instruments - rather than duplicating safety regulation . This complementarity is a deliberate design choice: even after everything is safe and all contracted value is delivered, externalities still accumulate from the myopic composition of many agents each focusing on their own business goals . She noted that safety dimensions might be incorporated into the framework as it matures, but that the initial commitment to make a deployment safe does not yet constitute the kind of "beyond business as usual" commitment the framework is designed to capture . Dr. Ulieru added that openness itself incorporates safety through collective inspection, arguing that formal models can be used to prove safety and that the platform's transparency enables the best AI developers to scrutinise it collectively .

The second audience question, from Sidoin Sotudo-Aiti of the UK Foreign, Commonwealth and Development Office, raised the concern of enforcement without a central mandate, asking how compliance can be achieved when actors fall out of line . Dr. Lenartowicz explained that the framework's enforceability is decentralised and does not require new institutions or certification bodies, because it operates through ordinary contract law already in place across human civilisation . By embedding commitments in standard legal documents - terms and conditions, API conditions, procurement documents, and sale contracts - actors create legally binding obligations that can be enforced by clients, civil society, and potential beneficiaries who can take the matter to court . She acknowledged that effectiveness depends on the jurisdiction and the strength of the surrounding social and civil society environment, but maintained that the legal system as it exists for contract law provides the necessary enforcement layer . Dr. Ulieru reinforced the safety-through-openness argument by contrasting open-source platforms with large AI labs operating as "opaque fortresses" where algorithms cannot be inspected, arguing that openness is what makes AI genuinely safe .

#

Broader Themes and Conclusions

Several overarching themes emerged across the discussion. First, there was broad consensus that existing AI governance instruments are insufficient and that a new approach targeting systemic effects across value chains is both necessary and timely . Second, the voluntary, bottom-up, and lightweight licensing design of the AIC Framework was widely endorsed as a viable and appropriate mechanism, with Creative Commons and open source cited as precedents demonstrating that voluntary frameworks can achieve significant systemic effects through self-organisation . Third, the inclusion of SMEs and individual commoners was identified as a central concern by multiple speakers from different institutional backgrounds, reflecting a shared recognition that the AI economy's benefits are currently concentrated among a small number of large actors .

A productive tension ran throughout the discussion between openness and controllability as competing safety paradigms, with Dr. Ulieru consistently championing openness as the more effective mechanism and Dr. Lenartowicz acknowledging both approaches . The internal disagreement within the working group about the Value Sharing pillar - with Dr. Lenartowicz proposing a UBI-oriented model and Dr. Ulieru preferring meritocracy, whilst tentatively acknowledging she may reconsider in light of AGI developments - was handled constructively and used to illustrate the framework's flexibility, though it also raised questions about the definitional coherence of the framework's standardised legal clauses.

Dr. Lenartowicz closed the session by inviting participants to continue conversations in corridor discussions and subsequent panels, noting that the working group would be bringing the framework to other sessions during the week . The session thus served its dual purpose of presenting the framework and gathering the feedback and expressions of interest needed to advance its development .

—
Dr. Em Lenartowicz
Hello everyone! I hope everybody can hear me in the room. Is it working? You can hear me? Yeah, okay, okay, yeah. But by the mic, yes? No? Can everybody hear me? Good. And I hope online participants as well. All is good, yeah. I just got thumbs up. Good afternoon and welcome to the session dedicated to the presentation of the AI Commons Framework, AIC Framework. This framework is designed with the intention to facilitate bottom -up self -organization across AI value chains, a self -organization that would amount to beneficial systemic impacts of the AI transition. So this is the focus of the framework and the intention. And the framework is being developed by a working group, which is housed at the Free University of Brussels. And you see five members of this working group in front of you. I will introduce our panelists when we move to the discussion session, just in a few minutes. And for now, I would like to guide you very briefly. through the logic, through the intention of the entire approach. This group that you see on the slide, the working group, Working on the framework is overlapping partially with the steering committee of the AI for Good impact initiative. So there are quite a few overlaps in those two bodies. And the framework is still under development, so this session is also intended to gather feedback and expressions of interest and networking. And I suppose when you hear a governance framework of AI, it's a phrase you have been hearing already for many times. It might be a reaction, oh, yet another governance framework. So the United Nations scientific panel has counted over 40 instruments intended at facilitating governance of AI. And the just released report that they have published is saying that, those frameworks are mostly fragmented, concentrated at the corporate level, and rarely measuring real -world effectiveness. So for what it's worth, the framework that you will hear about today is nothing like the governance frameworks that have been circulating and are being developed from the corporate angle. And the starting point is really real world effectiveness impacts, systemic impacts onto the world that AI systems, AI deployment, AI transition is bringing about. Not really in terms of measurements of those effects, but rather really inducing positive effects and ensuring that they propagate. That's the intention and the focus of AI. Of the entire architecture. How can we go about inducing beneficial systemic effects? And in terms of the corporate governance, of course, it's necessary and essential for corporations, for AI deployers and developers. to have assurance, to have safety in what they are controlling, that the systems they are developing and proposing are controllable. But the problem with approaching AI governance mostly on the corporate level is that, at least to my mind, there is an inherent problem with the business focus. Of most of the governance frameworks, in that they seem to be trying to ensure a similar level of controllability of business operations run through and by and with AI, similar to how humans behave and how human -led organizations behave. So the standard seems to be, OK, can we really make sure that the business run by AI behaves in a similar way? And yet... We all know that the human -led organizations have been creating many problems, and the standard of bringing AI economy to really ensure it behaves the way human economy does will not get us out of the problems we have been already witnessing and will only magnify them through acceleration. So really in the AI transition, we have to step onto higher grounds, learn how to do it and find ways, and enable the new types of organizations that involve AI and deploy AI to not create problems that human organizations have been. So that's the challenge. And the question and the design question or task of this... of this framework. And when you think about the exponential dynamic of how wanted or unwanted effects accumulate on an exponential rate, it doesn't happen merely through, it's not just aggregation. It's not addition of more and more projects doing something similar. The exponential dynamic doesn't happen through addition. It happens through multiplication. And what does it mean? It means that whatever dynamic we are seeing is emerging mostly out of interaction of the actors or units included in the cohort we are plotting the dynamics on this chart. So also in business, how does it happen that this very strong tendency to extract, to accumulate, to create externalities, to create exclusion. How does it happen? It doesn't happen through the individual choice and strategic decision of any single corporation in isolation. It happens through competitive dynamics, through the pressures that markets generate, through something through the realization that any actor cannot afford to behave different than any other. So it's this multiplication of tendencies that creates wanted or unwanted dynamic. And since we are able, you know, as human population on Earth to create unwanted exponential dynamics and we know we can, why not go about designing an exponential dynamic that is beneficial and in which AI actors actually pressure each other and expect from each other, operating on a higher ground and doing something that they haven't been doing before. So how can we establish such a practice and such a culture? The framework that we are working on is providing a shared surface and vocabulary, and basically vocabulary and grammar, you might say, that will enable just that, that invites any AI actor to formulate a commitment for the responsibility over aspects of AI operation that normally, in business as usual, so -called, would be kept outside. Outside of the boundary of attention or responsibility. So this is the primary descriptor of this framework. If you... So if you run an AI operation and you ensure it doesn't kill people and it's stable and it does what it wants, it's not yet the domain of this AI commons framework because, you know, like obviously you are supposed to ensure that, right? And also you are supposed to follow the law and regulation. But when you step outside of the boundaries and you make a commitment that is really impressive and really something new, you know, and in terms of taking responsibility for including externalities, including people that would normally be excluded, balancing your operations such that it's not extractive, not so strongly, aggressively concentrating all the gains, but it's sharing and inclusive and resourceful. Then you make this commitment through the AI common framework and the framework. is intended to really magnify this information and to let people know, let your customers, let the customer bases, let the society know, let your procurement partners know and policy makers that as an AI actor, you have made such choice. But a choice, individual choice, as fundamental as it is, doesn't yet amount to this multiplication effect that can create exponential dynamics. So the next move is to extend individual commitment of each project into a requirement across value chains. So not only you make a commitment, let's say for sustainability or openness or reciprocity, and I'll just walk you through the possible commitments which we are just starting to formulate as a library that can grow. So not only you take it on yourself, but you also expect that you will be doing business with partners downstream and upstream who actually commit to the same. And the framework allows you to express in a very legible way what are your systemic expectations from your suppliers upstream and what are the conditions of the re -usage or usage of your capabilities that you make available for operators that work downstream. And therefore, you can actually enter this dynamic of the mutual pressures, mutual expectations, and holding each other to higher standards. So on the compact level, on the top level, the framework looks like this. So it's a very small object intended to be readable by humans at a glance. Something like this. Like a QR code, but understandable. from just looking at it. And I understand that it's still cryptic to you now, but hopefully in three minutes it won't be, or four. Also in the corner you have an actual QR code to the web page where there's more details and more descriptions. So what you see, you see horizontal lines, those R -S -O -G -A -V. Those are types of commitments that any AI actors can make. So commitment for reciprocity, commitment for sustainability, openness, governance, access, open access to your system, and value sharing. And the verticals are where in your decision as an autonomous actor in the value chain, where this commitment lies. In the middle dot would mean that you take it upon yourself. This is the core commitment of your deployment. To the left you have expectations. upstream and to the right you have expectations downstream. So, for instance, if a project chooses voluntarily, it's like purely bottom -up and voluntary design, chooses to share a fraction of their gross revenues with broad classes of, broad groups of people represent, present, like aggregating people who are contributors to knowledge, to IP, you know, creators, to all sorts of data production labor. If your system finds a way to actually stream value, monetary value back to those groups, then you can, then you, and this is a formal legal clause that by putting this dot there, you can actually, you can actually, you can actually, you accept this clause. It's a similar formula. creative commons licensing or open source licensing, that by using an icon, this means you have accepted the standardized legal terms. What does it mean? What are your obligations for this aspect of responsibility, systemic responsibility for your social economic environment in that case? When you choose a commitment or mutual requirement, which is labeled as sustainability, and you could extend it throughout all your partnership, each of those profiles is a very ambitious choice. It's not business as usual at all, because S means, and it's detailed like that, that the environmental footprint, and cost, environmental cost of the entire operation within the boundaries of your deployment is balancing with the effects of this deployment, at least in a neutral way. ideally in a regenerative way. So, of course, not every project can do everything like that because not every project is regenerating the environment, but maybe every might, you know, once the system spreads. So you have projects like, for instance, I was just reading about the Chinese operation of robots, solar -powered robots that go to the desert and plant trees. You know, you can measure the environmental regenerative effect of this AI deployment against the cost. And when it's at least neutral, then the S is granted, and you can use this icon. The next one is Open S. It's an extension of open source. It includes open source that the code of your deployment is open, but not only. It extends to the operational level of your AI operation. And if you make your weights open, if you make your internal AI procedures open, such as incident handling, all things that civil society is concerned about, if it's scrutinizable by the society, then you put an O. And again, you can choose where you put your choice, where in the value chain around you. And the choice about governance is letting the full representation of all stakeholders groups for your deployment be present at the decision making table about the key aspects of the deployment. And I will just tell you one decision of this category, and it will tell you all. The sunset decision. The decision to actually close the system. If everybody can actually participate in the decision because it's harmful and you want to close it, and it's not only owners who decide that, but stakeholders, then the G applies. A project can take an A when they make a full set of capabilities, AI capabilities that are being made available through the deployment, free, in a mode of free access for everyone, meaning that it's free of charge for everybody. Again, not every project, but publicly funded projects, I suppose they should aim at this formula. And the last one is value sharing. This is more of a promise that the AI transition seems to be making of this universal abundance and this powerful acceleration of productive capabilities that AI is bringing, which should create material wealth. The question is, how do we share this material wealth? And if a project chooses V, like value sharing. This means that it's contributing to a universal basic income scheme. not running one but contributing and there are legal protections that the data is not shared and so on targeted at the region where the condensation of people living in an extreme poverty around the world is the highest so they can choose but there are criteria how to choose the region and of course if you are let's say a developer of a new model which will allow businesses run basically fully automated and you will see how it will change the entire business downstream, you could put such a condition downstream on your on your operators downstream why not consider that so this is the formula free choice free choice extending to reciprocal and and mutual pressures as as in societies we we tend to do and we have seen it happening before the voluntary frameworks such as Creative Commons and Open Source have already demonstrated that they can do wonders by sheer self -organisation so the question can it work of course we can ask ourselves about that but this is the question I want to address our panel first with and let's discuss the first person that we have scheduled for our panel is Ana Garcia Robles who is at Pal Expo at the moment in a parallel track so we have a recording from her and I'll play it and with us here at the table we have Dr. Anne Borda from the Alan Turing Institute. Everybody here at this table is a member of the working group working on on a framework. Then we have Dr. Christian Axenier from Nuremberg Institute of Technology. We have Dr. Michaela Uliero from SingularityNet. And we have Dr. Yang Yong from Hongking University. So let me first play Anna Garcia recording. She was just recording today, so it's very fresh.
—
Ana Garcia Robles
Hello, good afternoon. Thanks for the invitation to your session of the Commons Framework. And my apologies that I cannot be with you now. Well, as you can see, I will show a global dialogue in Palexpo. And it's very good that all these events have been organized together. But when we are not closely together, it's very difficult to communicate. between sessions and obligations and that's the reason why I cannot be with you today. First to thank Dr. M. Lenatovich for inviting me to be part of this important discussion that you are having in a few minutes. And just to say I'm Ana Garcia, I'm part of the United Nations Office for Digital and Emerging Technologies and I'm in particular in charge of a new lab on AI governance that has been established a few months back exactly to do more experimental practical work on AI governance, convening different experts and working with different networks, doing policy analysis but as I said also experimenting and looking at how we can find new innovative ways of governing AI both for the opportunities and for the challenges that it's raising. So that's the rest of it. And first in the AI governance commerce framework. And in recent work that we have just published, both in AI governance and probability and also in practices with the private sector, well, we have learned many things that I think what you frame resonates well. For instance, we have learned a lot about the shift in moving from AI governing more static AI systems to the dynamic nature of AI systems at the moment, mainly because of all the agents technology, and that's something that resonates well with your framework. Also the fact that AI governance is not only, it doesn't live only in norms and how those norms are implemented. But also it lives in a lot of decisions, interactions, it lives in value chains, in ecosystems. And there are many decisions that are made, for instance, decisions that customers make in procurement in different collaborations that are built in the ecosystem, and that resonates a lot with your framework. And also we are doing quite a lot of work in governance interoperability. The aspects of your framework, the six pillars of it also resonate a lot with some of the pathways that we have defined as a result of our work in governance interoperability. With that, well, it's just to transmit our desire to keep collaborating, to see whether there are opportunities to test, to pilot the framework, to learn more about it. Well, I wish you all a very good discussion today, and let's start the door open again. Thank you. Thank you very much. Thank you very much. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Thank you.
—
Dr. Em Lenartowicz
Hello, Dr. Anne is the head of the newly established AI Governance for Humanity Lab of the United Nations. So the lab is just starting now. And the next speaker at our panel, Dr. Anne Borda. Anne,
—
Dr. Ann Borda
Hello. I'm affiliated with the Alan Turing Institute as an ethics fellow. And for me, this framework is very important because of the value chain that's been discussed by M and Anna, and particularly the pathways that we would look through to see this materialize. This is effectively an operationalization of responsible AI principles, many of which you may have seen in the OECD principles, other recommendations that have come through other U of N agencies. So there is quite a lot of opportunity to see this. framework as somehow an operational arm to some of the responsible AI discussions that we have all been part of. And also in terms of engendering public trust, I mean, it's very important now where we see uneven regulatory landscape, we see a push to self -regulation, you know, where do we sit within that? And licensing as a sort of a pathway to make this viable is an interesting one because it's much more lightweight, but it's also much more part of an innovation cycle that governments are very much also involved in. The OECD has been very involved in licensing and permitting agenda. Which allows for those spaces between regulatory development. to be supported through licensing opportunities. Now, what does this mean in other spaces? So, for instance, the digital public infrastructure is one area where we can see this hanging onto in terms of looking at DPI and channeling public procurement, for instance, towards responsible AI and services, and particularly by limiting the ability of larger corporations to consolidate power through vertical integration. So what we're looking at is really opening up options beyond what we can see now. Now, you might ask, well, we've talked about open source as maybe a licensing example. Creative Commons is another. You may be thinking about the licensing arrangements where we have much more free choice. Open source has permissive. It's a copy lab. that's a very open license. We have much more restrictive enforceable ones, Creative Commons, for instance, and these are all being also very much shaped by the AI ecosystem. So we really do need to think about licensing as a pathway to also how do we improve licensing arrangements, particularly if we're thinking about companies and dual licensing, for instance, where you may have have licensing could become a checklist for one thing, but you're monetizing with another type of license. So this is an interesting conversation to have. My other colleagues will give some examples, but there has been precedence in social license to operate. If anyone has heard of a social license to operate, it's gone way back in time when mining companies were coming in and communities were building. up a social license to operate. So there was this dynamic between communities and companies. And we can see this framework as a sort of social license to operate. So a very exciting space, lots to think about and discuss.
—
Dr. Em Lenartowicz
Thank you so much, Anne. Thank you. And Christian, actually, please take it away.
—
Dr. Cristian Axenie
incorporated that in their business model to small agile companies. Well, at least in Germany, we have a lot of small, medium enterprises, and they're looking at that right now. And they need a way to actually give this added value to their customers. And there are examples already which already deliver more than a product, a system. They deliver a certificate on the product. And I think right now in the current context, this plays a huge role because this shows that we make steps towards harmonizing the whole deployment perspectives. And AIC tries that. As you see already, we are coming from different backgrounds, different perspectives. I think it pays off. This pays off. And just for like two more points, one is awareness. A lot of small, medium companies are aware that AI can bring value, but they don't have yet a clue or a recipe on how to actually adopt that. So what we're right now trying to push is the awareness and acceptance, because then adoption will concur. And now looking also at what Em mentioned with this velocity, this exponential, we know from mathematics such functions grow pretty fast, but they're also faster functions. Also people like having a bit of mathematical affinity, I see them laughing right now. This is important to understand that in order to tackle that or to actually grasp that, to handle this pace of AI growth, we need to also have measures which grow at least as fast with the AI systems. And I think this this nice growth curves perspective should give us an understanding that we are here to do something which is pragmatic. And I think this like openness, sustainability, access and value, overall reciprocity. of the AIC framework gives us at the first hand the tools and second the means to actually pragmatically adopt that and we are all advocating that in our communities and I think right now we are on a good path to also make also here in this special setup awareness
—
Dr. Em Lenartowicz
Thank you so much Christian Can you please show
—
Mihaela Ulieru
Thank you so much Em and thank you for introducing this amazing framework I'm going to talk to you now about a real life project that I consider to be the poster child for the framework so let's see how it applies in real life and this is SingularityNet which is the only open source decentralized artificial general intelligence platform for with the people that reaches everyone So it is accessible. It is open source. We can call it the Linux of AGI, artificial general intelligence. So that's why I asked Em to show again, yes, all these dimensions, because I'd like to refer to them. So obviously, yes. So the project has been around since the early 2000s when Dr. Ben Gertel and a group, I'm a professor of decentralized artificial intelligence. So Dr. Ben Gertel and a group of research scientists have started this movement of artificial general intelligence open. It is now our Hyperon platform. So it is open source. Everybody can see what it does and everybody can access it. However, how do we do the reciprocity? Well, that is the decentralization part. And we also have deployed the only artificial general intelligence. It's called artificial super intelligence chain, ASI chain. And what is its role? Number one, your data, yes? I mean, you can have the cake and eat it too, because the data can stay private, but then encrypted via blockchain technology, advanced this AGI native blockchain, you can put the data in a pot together with others in order to get much better results for whatever you do. And therefore, but not only much better results, which would be already a sort of reciprocity, yes? But also being paid for the data which is involved in a certain product. On the other side, on our platform, you can post your own algorithm. And then, for example, if I post an algorithm and M is using it, automatically I'm going to be paid again via the blockchain and encrypted technologies. In terms of governance, yes? Well, obviously, decentralized governance in which everybody can. Can contribute. So we actually are doing that through a constitutional collective based on reputation. And that is commensurate with the expertise. So, for example, if M has much more expertise in governance than me, then I'm going to delegate to her to make decisions on my behalf because my weighting is too low. And vice versa in my case for artificial general intelligence, let's say. Then, well, one thing which is very important here is also the sustainability. Why? Because our computing power is diffused across the network without a need for the big data centers. Yes. So we have the data distributed, as I mentioned to you, through this blockchain, but also the computing power. So the sustainability is really achieved here. One thing which I think it's a caveat here, and I had big discussions with M about value. We see value a bit different. We see it as meritocracy. And that means a value commensurate to the contribution. And therefore, instead of UBI, we are organizing hackathons in which people can contribute code and really be rewarded for real work. Because we really don't believe in doing nothing and getting money back. However, now with the change that AGI is having, including my own as a professor of AI, they start to become smarter than me. I may be reconsidering the UBI. Thank you.
—
Dr. Em Lenartowicz
Thank you very much, Michala. And what you say about this V and not wanting this, I think this is a wonderful illustration of the entire bottom -up and voluntary aspect of it. Because this framework allows you to express what you believe in, what you actually want to contribute. It's not that it imposes that you need to do that or that. You choose, and your environment actually checks if they are attracted to this choice of value or not. And just one last thing. With the value chain, we have the artificial superintelligence alliance where we apply or can apply this upstream -downstream value chain. Yeah, yeah. system. And Dr. Yang, you'll please take it away.
—
Dr. Young Yoon
Okay. So the group that I'm more interested in is the commoners, individual commoners and small to mid -sized businesses that are left out of this AI economy. And I think we have to get down to the fundamental issues faced by these individuals and SMBs when it comes to participating in this new waves of AI value chain that is heavily dominated by big tech and also just a small number of technocrats. But for these small players, the individuals and SMBs, they're not going to budge unless they are incentivized with a sizable reward that is given upfront. Right upon the contribution of their personal data, for example. But it's very difficult to reward them. because it's going to take time to create value -added AI services with these personal data. But this is a human nature to be very impatient. And also, these small players are also very concerned about the breach of their private information. And this feels very ironic to me because all these small players already give full consent to the big tech to share all their personal information while getting no compensation. Even so, to solve this problem, I think one way is to create a cooperative with these AI commons so that... These cohort of shared business interest can gather together and... share all their private resources, and create shared common value -added services. And one way to eliminate any concerns about the security, we can use distributed learning algorithm, where you can circulate the AI models through different sources of private information without exposing any of the sensitive data. And once all these AI models are created, then you can go for profit sharing. But even with this framework of cooperatives, it's still very difficult to reward all these individuals right up front. So I think we should start with small to mid -sized businesses. And then once we learn all the principles to make this work, then I think we can extend it to the individual. I think that's the right pathway and more pragmatic path. So that's my two cents.
—
Dr. Em Lenartowicz
Thank you. And we have five minutes left, so I suppose two, three, maybe four questions, depends how fast we are. Yes, please.
—
Audience
Hello. Thank you for a very interesting session we are having right now. I have a question regarding, first of all, licensing, then the last thing Dr. Young was talking about, about sharing revenues, and Dr. M also talked about the revenue because you were committed on developing something. But what is the other side of the responsibility? If the model is something built, made some error, occurred some harm, who is taking the responsibility then? And should this be included in a licensing process? I mean, we already have EU AI Act and the similar legislative acts in Afghanistan. There are states and countries, and they all are talking about how to govern. But today we do not have any tool, any standardized tool, to make it simple and understandable for any state, any jurisdiction, and any type of industry. Thank you. so how do we include safety and those concerns I mean yeah how we should include the safety and make it in a standard way to be applicable to any kind of industry any jurisdiction or state and how it should be managed besides the rule maybe we should talk about having some templates platforms or like a financial reporting system what we did far ago but it is a standard for every industry or we have a nutrition label for the food and it can stand as a standard what do you think about that
—
Dr. Em Lenartowicz
so so it seems to me and this is this is more of a private private view and and we we might add as our working group proceeds this this type of safety assurance and and other aspects that are very heavily discussed currently but it seems that the majority of of of the AI governance instruments and also regulation, of course, as, you know, undone yet it is, and the discussions are ongoing, are actually focusing on how do we make sure that the deployments are safe and lawful and so on. And, yeah, we don't have a world governance, so depending on in which jurisdiction you live, the approach will be different. And the idea was more about, in this framework, with complementarity, which commitments and which aspects of the accumulating systemic effects are actually falling through the cracks. So it's a little bit different focus because the commitment, I will make it safe, it doesn't seem like to me, you know, but we might still include safety. As something like beyond business as usual commitment, it is everybody's duty to make sure that what you are bringing, to the market is safe. safe and regulation needs to be able to execute this type of compliance which of course is the work in progress and we don't have it and after everything is safe and after what you are committed to is delivered we still face those aggregating systemic effects just coming from this myopic composition of multiplicity of agents each focusing on their business goal and delivering value as contracted and yet externalities accumulate and so on so the focus is a little bit different but you know once our framework becomes established we might factor in different levels than only this initial set. Michal I want to continue.
—
Mihaela Ulieru
Yeah I just briefly want to say that the framework in itself contains the safety and I will explain how because we do this in real life as a professor of AI we actually use formal models to prove the safety but on the other side the public on chain Expectable cognition. This is what we have with our platform, which is open source. So we have all the eyes of the best AI developers in the known universe on the platform. And then you can really see. So through this openness, you already have the safety. It's the best immune system on order, on offer today, because this is transparent, decentralized, interpretable, diverse network, which is collectively defended. So that already has that incorporated. And then you can have that automatically. You mentioned the food and so on. Well, with AI, you can have that automated now in order to have that kind of inspection through formal models. Anyway, we can talk about that.
—
Dr. Em Lenartowicz
And openness is obviously one of the two possible approaches to safety. Another is controllability and auditability from the top down. But bottom up is something that you can already claim through this. Oh, as openness profile, we can we can still take one question and it will be like time to.
—
Audience
If no one else has, may I continue?
—
Dr. Em Lenartowicz
There are a few other people. You were raising your hand. Yes, please.
—
Audience 2 - Sidoin Sotudo-Aiti
Yeah, thank you very much. My name is Sidoin Sotudo -Aiti. I work for the UK Foreign, Commonwealth and Development Office. Excellent presentation, first of all. I think we see the problem of 40 regulatory instruments around AI, but yet we still have a challenge. That's why I feel the framework may actually solve the problem. My immediate concern would be on the centrality of the mandate, the point you made about where there are disagreements, discrepancies, or someone just falls out of line. How do you enforce? Because in that instance, you would need a central mandate to be able to do that. That is my primary issue here.
—
Dr. Em Lenartowicz
Yeah, so this is the primary feature of this design, the enforceability, which is decentralized and actually doesn't require new institutions, doesn't require new certification bodies. nothing like that, because it works through ordinary contract law, which is already in place everywhere, and the whole human civilization relies on that. Because once you reference your commitment in your regular documents through which you make your deployment available, terms of conditions, API conditions, procurement documents, sale contracts of your deployment, this is how you bring in your commitment as a license, because you write those licensing commitments into those documents, and then there is another party who can sue. So, of course, depending on how many eyes are looking at your commitment and how many stakeholders are there, but you have legally, contractually obliged yourself to the particular shape of your deployment. Then your client. Then public society. Potential beneficiaries that you promise to benefit, and you are not. they go to court and the enforcement layer is already in place the legal system as we know it for contract law so that's the idea and again depends on in which jurisdiction how effective that is and how strong the environment around a particular deployment, the social environment and civil society environment is but that's the angle of the approach. Unfortunately our time is ended so let's take it to the corridors in the coffee breaks and so on yeah go ahead
—
Mihaela Ulieru
I just wanted to say one more thing which I think is important because people are not used and this is to your question to what is openness open source and why because artificial intelligence is now developed in the big labs and that's what we know as an opaque fortress with the lights off and we do not know what mysterious powerful algorithms are inside we cannot inspect that But with open source, everybody can see the code and can see if it is and evaluate if it is safe or not. And, you know, just because those walls are thick, they doesn't make AI safe in this case. But openness and open source is what actually the power of this. I just wanted to make this contrast because most people are used to GPT, open AI, but those are the closed fortresses. That's why they are not safe.
—
Dr. Em Lenartowicz
Yeah. Thank you so much. And anybody interested in continuing the conversations, let's find each other. The week is only starting. We'll be we'll be bringing this to other panels as well. So. So see you. See everybody around. Thank you.

Disclaimer: This is not an official session record. DiploAI generates these resources from audiovisual recordings, and they are presented as-is, including potential errors. Due to logistical challenges, such as discrepancies in audio/video or transcripts, names may be misspelled. We strive for accuracy to the best of our ability.