top of page

The Argument for Co-ownership of Agentic AI

Explore AI Agents today, and join the rallying cry

Amid great advances in AI, Artificial General Intelligence looms and poses immense opportunities and great risks for humans. Valory believes AGI will likely be agentic and in all cases must be co-owned to avoid devastating outcomes for societies. Blockchains laid the foundations for decentralized ownership of digital resources and applications, and now Olas (previously Autonolas) is enabling co-owned AI agents that you can experiment with today, including co-owned AI DAO governance, prediction agents, and much more. Join a rallying cry across decentralized AI and open-source AI communities.

AGI (artificial general intelligence) will likely be agentic, and must be made co-ownable.

Fig. 1 Artistic illustration of a co-owned, agentic AI system.

Today, AGI is still a theoretical concept, but some of the world’s most accomplished AI companies, including OpenAI and DeepMind, are on a mission to create AGI. In the case of OpenAI, the mission explicitly aims ‘to ensure that artificial general intelligence benefits all humanity’.

Whilst this is a laudable mission, it is only a mission right now. There is no forcing function that ensures that, should OpenAI or another company ever develop AGI, they will indeed deploy it in a way that ‘benefits all humanity’*. Recent events at OpenAI highlight that there are severe limitations to designing governance systems for AGI development that rely on traditional legal forms such as companies and non-profit organizations.

As it is still a theoretical construct, it is not guaranteed that AGI will indeed be agentic, i.e. make independent decisions and take actions based on its own goals or motivations. However, optimization pressures suggest that AGI will likely have a level of autonomy and self-direction that goes beyond just being generally intelligent. The recent proliferation of AI agent frameworks, and the huge benefits that can stem from agentic AI, suggest that at a minimum we will have highly effective agentic AIs.

It seems evident that with increasing advances in AI technology and the power that these advancements bestow on their owners, the future of humanity is increasingly at risk not only from the agentic AIs themselves but also those who control them. Whilst some argue for stopping these advances altogether, we do not think this is feasible**. Instead, we must focus all our energy on ensuring that agentic AI, and ultimately AGI, do not have central points of failure, but rather can be co-owned and co-controlled on a technical level, too.

Without true co-ownership of powerful agentic AIs, we risk creating at best extreme economic disparities that destroy the fabric of our societies. Conversely, Valory foresees and is working towards a future where true co-ownership of AIs (both pre-AGI and AGI level) will be key for the flourishing of every human, organization, and country.

Blockchain laid the foundations for co-owned AI

Since the arrival of Bitcoin in 2008, a fast-growing number of protocols endeavor to decentralize control over critical resources, like currencies, data storage, compute, and others. Whilst the mechanisms and ambitions vary widely, from thinly veiled scams to serious global hyperstructures, it is undeniable that they have given rise to novel forms of coordination and control structures that use code instead of or in addition to law***.

Increasingly, protocols are being developed to decentralize various aspects of the AI stack, too. This stack has many layers**** and is still in its infancy*****:

  • Ownership (IP & Revenue)

  • Operational Control

  • Development

  • Training

  • Deployment (Inference & Serving & Actuators)

The more open, transparent, and auditable these layers are, the more humanity can put real checks and balances in place via appropriate governance systems that are enforced via code, cryptography, and game theory. Likely these governance systems will have to lean on the systems that they govern to be effective: tools of the autonomous age are needed to control tools of the autonomous age.

Olas is enabling co-owned AI

At Olas, the founding members started with the humble goal of building a stack that allows anyone to co-own an off-chain autonomous system. In the crypto space, this has immediate applications for Decentralized Autonomous Organizations (DAOs), allowing them to take full control of their off-chain stack.

Fig 2. Olas is an ocean of agent-powered services – the substrate for co-owned AI.

However, what the Olas ecosystem is creating goes far beyond crypto. We are creating a substrate for co-ownable agentic AIs. This is what we at Valory have been increasingly focused on since earlier this year. Our co-owned agents take advantage of the emerging decentralized AI stack for model training and inference, as well as various open- and closed-sourced AI tools.

Fig. 3: Autonomous AI Agents Tech Stack

Governatooorr, an autonomous AI agent participating in DAO governance on behalf of humans, is a great example of this, in its infancy******. The AI agents participating in prediction markets on Gnosis Chain are another*******. Both these and other products the Olas ecosystem has launched provide early proof points of our mission.

Fig 4. Governatooorr and AI Prediction Agents – early examples of co-owned AI agents

Experiment with co-owned AI agents

Clearly, AI will continue to rapidly evolve and have an outsized impact on human life. It is vital we explore ways to ensure the benefits and risks are managed by society at large rather than a select few. Given agentic AI is the most likely path to AGI, it is an exciting time to experiment with autonomous AI agents. They sit at the nexus of three key developments in AI, web, and UX. There is a tremendous amount of work to be done to create increasingly capable autonomous agents that are co-ownable, safe, and aligned enough to be adopted en masse********.

Today you can contribute by building agents, improving the tech stack, contributing research, and more.

Join us, making waves!


* Let us ignore for the moment the fact that it is not even clear how to measure the attainment of such a goal.
** There is a possibility that there is only one non-aligned AGI and we all die. In the absence of this scenario, there will be many agentic A(G)Is competitively interacting (like humans do).
*** On code is law.
**** For a more comprehensive discussion of the tech stack see and Messari.
***** There are multiple challenges around technical feasibility, scalability, and security concerns that need to be overcome (e.g. 1, 2, 3, 4 and 5).
****** For a critical examination of Governatooorr consider this piece.
******* Trading agents on Gnosis are generating more than half of Safe transactions monthly for the second half of this year.
******** From distributed systems and consensus mechanisms, across game theoretic and mechanism design, all the way to the alignment and ethical challenges.


bottom of page