*By Marcelo Almeida
Bill 2338/2023 provides guidelines for the governance of artificial intelligence (AI) systems in Brazil, and brings a new horizon of opportunities and challenges.
Small and medium-sized national entrepreneurs tend to be more impacted. This is no different when it comes to artificial intelligence. Therefore, throughout this article we will address five points of attention that should be considered in practice when promoting regulations on artificial intelligence applied to micro and small national entrepreneurs.
The first concerns governance and transparency. Bill 2,338/2023, which is currently being processed in the Federal Senate, requires AI developers to maintain detailed documentation on how the systems work and the decisions involved in their construction.
The requirement for detailed documentation from micro-entrepreneurs can be challenging. Creating and maintaining this documentation requires financial and human resources that are often scarce. Transparency can increase consumer trust in AI products developed by small businesses. The cost, however, can be mitigated and distributed across the AI value chain to lessen the impact on small and medium-sized entrepreneurs.
A second point concerns algorithmic impact assessments, which will be shared with the competent authorities. These assessments involve identifying risks and benefits to fundamental rights, as well as mitigation measures.
Conducting algorithmic impact assessments requires specialized knowledge that is time-consuming and costly. For small entrepreneurs, this means hiring specialized consultants or investing in in-house training.
The third aspect concerns the increased cost of implementing internal security processes throughout the entire life cycle of systems. This includes the use of tools for automatically recording system operation and carrying out reliability tests.
Implementing security measures also presents a technical and financial challenge for microentrepreneurs. Most microentrepreneurs do not have the knowledge to, for example, integrate systems, or even to choose instrumental priorities for the results of their own business. Printing the responsibility may mean raising the quality standard of the products developed, making them more competitive in the market, but it may be a requirement beyond the reach of small entrepreneurs.
A fourth point concerns the provisions of Bill 2338/2023 regarding measures to mitigate and prevent discriminatory biases and encourage diversity in development teams. Management and governance policies are also needed to promote social and sustainable responsibility.
For micro and small entrepreneurs, implementing diversity and social responsibility policies can seem like a challenging task. It is common to find a closed and resistant organizational culture. Many employees work throughout the production chain. Imposing new procedures, training, and integration will require a great deal of convincing, which is costly to the business. Diverse teams tend to be more innovative and effective in identifying and mitigating biases. Adopting socially responsible practices can also improve the company's image and attract talent and customers who value these principles. Overcoming the implementation barrier, however, can be complex.
Finally, developers must ensure the explainability of the results of AI systems and provide adequate information that allows the interpretation of these results, respecting industrial and commercial confidentiality.
The guarantee and explainability of AI systems require a great deal of investment in development and communication. Micro and small entrepreneurs may need to create more intuitive interfaces and provide clear information to users. By the way, making AI explainable may mean opening a black box of machine learning that composes it, compromising business singularities and differentials. Anything that is explainable is popular with consumers, especially small ones that demand more trust from them compared to large, established brands in the market.
Bill 2338 presents a set of challenging governance guidelines. Making governance elements mandatory can bring benefits to Brazilian micro and small entrepreneurs. The issue is the cost of indiscriminate implementation, without adequate modulation for businesses of different sizes.
Having detailed and transparent documentation, promoting algorithmic impact assessment, security measures, promoting diversity and explainability of AI systems are elements that require investment, training, cultural change, knowledge, among other skills that only large entrepreneurs can hire in the market, whether due to cost or availability. If well implemented, governance mechanisms are a candidate to raise the quality and competitiveness standard of the products developed. However, establishing vertical, binding paradigms, with more rights and duties than incentives for the adoption of AI as a technology is a challenge that can be costly, reduce opportunities and make it difficult for microentrepreneurs to position themselves in the technology market.
*Marcelo Almeida is Director of Government Relations at the Brazilian Association of Software Companies (ABES)
Notice: The opinion presented in this article is the responsibility of its author and not of ABES - Brazilian Association of Software Companies
EN 
PT 
