A neutral and unbiased AI governance and safety system for the world

Project summary

I'm a researcher in the humanities with a background in engineering and programming. About a year ago I became interested in AI safety and started reading papers, technical reports and books about AI governance, AGI and autonomous AI systems.

I expected to find many technical proposals for how advanced AI systems could be governed once they begin making decisions and interacting with the outside world. Instead, I found that much of the work focused on describing the risks or discussing regulation. Those discussions are important, but I found relatively few proposals for a general technical governance layer that could work across different AI models and agents.

I've always enjoyed trying to solve difficult problems, so I decided to explore whether such a system could be designed.

During the last year I've worked almost entirely on that question. The result is an architecture that sits between AI systems and the actions they perform. Rather than replacing AI models, the system is designed to govern them by enforcing policies, authorizing actions, maintaining audit logs and allowing human intervention when necessary. The architecture is model-agnostic and is intended to work with different AI providers.

To protect the work while continuing development, I've filed four patent families that have now entered the international PCT phase. I'm also discussing pilot projects with universities, companies and other organisations. Internal testing has produced encouraging results, and the next step is to validate the system in real-world environments.

I'm also the father of two children. Like many parents, I think about the kind of AI society they'll grow up in. That has been a strong motivation for spending the last year working on this project.

What are this project's goals? How will you achieve them?

The main goal is to move from architecture and patents to working pilots.

The first objective is to continue the international patent process. International patent protection is expensive for an independent founder, but I believe it's important if the technology is going to remain available for broad adoption rather than ending up under the control of a single organisation.

The second objective is to build MVPs and pilot projects together with external partners. Testing the system in real environments will generate both technical validation and practical feedback for further development.

The third objective is to continue collaborations with universities, companies and public institutions while gradually building a small team around the project.

How will this funding be used?

The funding would mainly be used to develop MVPs and pilot projects.

It would also provide enough runway for me to continue working full-time on the project for approximately six to nine months.

Part of the funding would cover international patent costs, early collaborators and travel to meet research, industry and funding partners.

Who is on your team? What's your track record on similar projects?

At the moment I'm leading the project myself.

My background combines engineering, programming, the humanities and social sciences. I believe that combination has been valuable because AI governance is not only a technical problem but also an organisational and societal one.

Before this project I founded several startups, including an environmentally focused search engine and a sustainable Champagne company. I've also worked with technology and cybersecurity-related projects.

The project is currently supported by advisors with experience in technology and startups, and I'm looking for additional collaborators as development continues.

What are the most likely causes and outcomes if this project fails?

The main risk is insufficient funding before the first pilots can be completed.

If that happens, research on AI governance will continue, but I think independent governance infrastructure may develop more slowly. Most solutions are likely to be developed inside large AI companies or governments. I believe there is value in also exploring independent approaches that are not tied to a specific model provider or national interest.

How much money have you raised in the last 12 months, and from where?

Approximately USD 15,000, from family and friends.