The speed and scale of artificial intelligence evolution are transforming societies and economies globally. As these rapid changes challenge traditional regulatory frameworks, innovative solutions like sandboxes are emerging as essential tools to foster experimentation while maintaining safety and accountability.

Sandboxes are controlled environments where new technologies and regulatory frameworks can be tested safely. These spaces allow for experimentation and learning without the full risk of deployment, ensuring both technological advancement and regulatory compliance.

Initially introduced in the fintech sector, sandboxes have now expanded to other domains, including AI. As explained by Lorrayne Porciúncula, Executive Director of the Datasphere Initiative,

“sandboxes offer agile and innovative methods to unlock the value of data responsibly while ensuring trust and collaboration between stakeholders.”

On December 2nd 2024, as a side panel at the Open America – Connection without Borders the Datasphere Initiative hosted a session bringing together experts from diverse sectors to explore the role of sandboxes in navigating the complexities of AI. Here’s a summary of key insights shared during the event.

The Growing Need for Sandboxes

The rapid pace of AI innovation has highlighted gaps in traditional regulatory processes, which are often too rigid to adapt to emerging technologies. Fernando Maciel, Regulatory Modernization Coordinator at Brazil’s Ministry of Development, underscored the importance of trust and transparency in sandboxes:

“Sandboxes are spaces where the regulator works together with the private sector. The pillars of these spaces have to be the building of trust and transparency. There’s no way the regulator can effectively implement a sandbox without social participation and transparency of information.”

Cristina Elsner, Task Forces Employment & Education, Digital Transformation PMO at B20/CNI, emphasized that traditional models of regulation are no longer viable:

“The old model of regulation no longer works, we need flexibility with security.”

Cristina added that AI is not a sector, it’s a technology urging policymakers to consider AI’s diverse sectoral applications and inherent risks. She explained that we need to think algorithmically, think about AI through its sectoral application and at a risk level that needs to be experimented with.

Key Challenges for AI Sandboxes

The panelists outlined several challenges and considerations for implementing effective sandboxes in AI:

1. Sectoral Nuances: Sandboxes must be tailored to the specific dynamics of each industry. Cristina Elsner explained this by likening public policy pilots to real-life experiments rather than controlled lab tests. In these pilots, the lab test subjects are the institutions that are producing technology, such as schools, universities, health, and research centers. To ensure success of the sandboxes, all participants must agree on clear rules governing the process.
2. Legal certainty and governance: Fabiana Cebrian, General Coordinator of Technology and Research at Brazil’s National Data Protection Authority (ANPD), highlighted the ANPD’s commitment to testing Article 20 of the Brazilian LGPD (General Data Protection Law). The article specifies that the data subject has the right to request a review of decisions taken solely on the basis of automated processing of personal data affecting his/her interests. In a controlled environment, sandboxes allow stakeholders to experiment with transparency mechanisms for notifying individuals about the impact of automated decisions, design efficient and fair review processes, and ensure AI systems are explainable and compliant with data protection laws.
3. Collaboration among stakeholders: Effective sandboxes require multi-stakeholder engagement. This was emphasized by Raíssa Moura, Global Data Protection Officer and Head of Digital Affairs at Nubank, who stressed the importance of bringing different players together to ensure a productive sandbox environment. In her intervention she warned that progress is impossible if stakeholders remain isolated in their own silos without meaningful communication.
4. Cybersecurity and inclusivity: Sandboxes must address the unique needs of small and medium-sized businesses, which are often overlooked in regulations focused on major players like Google or Microsoft. By creating controlled environments that prioritize cybersecurity and inclusivity, sandboxes can provide these smaller enterprises with the support and protection needed to harness AI’s potential, enhancing their productivity while ensuring compliance with regulatory frameworks.

A Vision for the Future

The event highlighted successful sandbox examples, such as those in the financial sector (PIX, open finance) and initiatives in countries like Singapore and the United Kingdom. These models demonstrate how collaboration between regulators and private entities can result in cutting-edge innovation and inclusive public policies.

The rapid advancements in artificial intelligence call for a transformative approach to how regulation aligns with innovation. In this context, sandboxes present a vital opportunity to close the gap by creating collaborative spaces where governments, businesses, and civil society can work together to design ethical, efficient, and forward-thinking AI applications. By adopting and leveraging the potential of sandboxes, countries like Brazil have the chance to transcend its role as a mere consumer of emerging technologies and establish itself as a global leader in AI innovation, shaping its trajectory rather than simply receiving its outcomes. 

Learn more about sandboxes on our 5 minutes introduction on sandboxes  and the Global Sandboxes website. If you are interested in following news on data governance and the activities of the global network of the Datasphere Initiative join us as a partner or a friend and follow-us on LinkedIn.
The post Open Americas Conference Side Event on Sandboxes appeared first on The Datasphere Initiative.