By Sophie Tomlinson, Deputy Executive Director, Datasphere Initiative
Last year, I traveled to Uzbekistan to deliver training on AI sandboxes as part of a project to support “Adoption of AI in the Private Sector,” in partnership with StratejAI, the UNDP and the country’s Ministry of Economy and Finance. In addition to the training, my goal was to examine the various ways in which sandboxes could assist in the development of Artificial Intelligence (AI) technologies to support the growth of micro, small, and medium-sized enterprises (MSMEs). The key metrics and findings are included in the report, but the visit and the work taught me a few things that go beyond what was reported, like the importance of understanding the uniqueness of the whole country’s economy for sandboxes to deliver an actually diverse positive result. Reflecting on this experience made me think about how much other countries could get inspired by the Uzbek experience. In this article, I’d like to draw on my experience and technical expertise to share some insights on the matter.
Why AI Sandboxes?
Getting started with my research showed me that Uzbekistan is swiftly becoming a strategic AI investment opportunity in Central Asia. The country’s National Strategy for the Development of Artificial Intelligence Technologies until 2030 is a stimulus for economic diversification. With a view to implementing AI as a truly secure business development technology, one that benefits not only companies and the government but also the people of Uzbekistan, sandboxes could become a strategic platform for Uzbek MSMEs.
AI sandboxes are the controlled environments where MSMEs can test AI innovations with regulatory oversight. They can also offer access to data and compute as well as the technical support they need to ensure security and legal compliance without major financial impact. They often provide operational testing infrastructure to help companies develop and validate AI applications before full market deployment.
Since regulations on AI are still being built and negotiated in several countries, sandboxes become a particularly valuable way for navigating the uncertainty of the technology. The access to data, compute resources, and testing environments is usually a technical infrastructure that MSMEs cannot afford. Sandboxes not only help on that front, but also provide consumers and investors with validated proof of concepts in controlled scenarios, which gives MSMEs market credibility to keep evolving the technology and developing their businesses.
The challenges sandboxes can resolve
My time in Uzbekistan revealed to me the unique challenges faced by MSMEs when developing AI products. Some of these challenges can hinder their implementation of new technology, and some can be addressed by sandboxes. Clarity about business and regulations, for example, is key for these companies, but the interpretation of regulation is not always unique, precise, and uniform. Even when regulations exist on paper, understanding how regulators will actually interpret and enforce them requires profound knowledge and experience with the legislation, a service (and skil) that usually requires a lot of time and study and is only provided by top law firms with the highest fees. MSMEs often cannot afford this extensive legal consultations.
Many MSMEs lack access to sufficient data sets for training and testing. Synthetic data can be used in the early stages of a sandbox, for initial testing and development. It can be effective, but it is expensive, and the process may need to be supplemented with real-world validation before full deployment.
It’s not only about the data inside the sandboxes – it’s also about the humans getting together during this process and their interest in developing safer and more useful technology for all. I’ve seen this in Uzbekistan, but also in every opportunity I have had to monitor similar environments for sandboxes.
The particularities of AI sandboxes and why they are very helpful for developing economies and businesses
AI sandboxes are particularly important for countries struggling with regulatory uncertainty, which is not a farfetched reality, since many countries are still developing AI governance frameworks, or monitoring existing regulations’ ability to cover AIapplications. The EU AI Act, for example, is also facing challenges with the simplification agenda because AI has evolved so rapidly that aspects of the General Data Protection Regulation (GDPR) have become challenging to implement in practice.
In such uncertain scenarios, most countries and businesses around the world are building on the unknown. Experimentation sandboxes can be really helpful for that, not only because they are a practical way for deploying new technology, but also because they are a strategic way for fomenting dialogue among innovators, fostering debates on AI applications, and building technology with accessibility in mind. Therefore, MSMEs can use AI to test proof-of-concepts while simultaneously including the communities that are going to be impacted by the tech they are developing.
Since the level of risk from an AI application can often depend on context: how, where, when, and who is using the tool. This is what makes real-world testing in controlled environments essential for understanding actual impacts rather than theoretical risks. It builds consumer and investors confidence, because if your product has been through a sandbox, it means it has already been tested.
AI is not only for China and the US
When I visited Uzbekistan, the sandbox initiative was still in early stages, so the full implementation and results remain to be seen. However, one thing that became quite clear to me is that investing in Artificial Intelligence advancement is actively on the agendas of countries around the world, and not only to the nations at the center of current debate, which focuses on the AI race between the U.S. and China. The gap in private sector investment between the US and other countries remains massive, but many nations are developing national AI strategies and looking for ways to foster domestic innovation.
On that note, Uzbekistan provided to me a specific insight into how a country outside of the hegemonic AI debate can approach the technology through sandboxes. The sandbox concept offers a way to connect the country’s innovation ecosystem. Universities, business associations, incubation hubs, and the government are exploring how they can collaborate around concrete testing projects. All of these entities normally act separately, dividing innovation into a traditional and not agile method of tech advancement. Instead of building more silos, sandboxes cab build a collaborative space to foster inclusive innovation. For developing countries facing challenges related to investment and representativeness, it could be strategic and more economically viable.
A Perspective of advancement in a global scale
In conclusion, from all the lessons I’ve learned from my visit, the most valuable one is probably how sandboxes are a very helpful tool for the development of MSMEs.
They can also bridge gaps between key stakeholders that would not necessarily collaborate. This encounter of perspectives, experiences and expertises are vital for the brewing of ideas that can build new paradigms of quality and efficiency for MSMEs. This is how new products and services brought by the exploration of AI advancements will have a positive impact in the day-to-day lives of huge population numbers.
For the rest of the world, the Uzbek experience becomes a lesson: AI is not just chatbots developed by rich countries with vast amounts of money to spend. The application range is immense, and with sector-specific requirements, AI use case can be seen in agriculture, transport and health-care. That is why context matters a lot, not only for tech design or creation, but also for risk mitigation and reality checks. In that context, sandboxes can emerge as very helpful tools, assisting governments to understand what regulations support responsible innovation and eliminate potential harms and risks, while businesses can focus on developing new technology in a safe and collaborative environments.
