*By Cibele Perillo
I've always liked poetry, especially haiku (or haikus), which are short, three-verse poems of Japanese origin. For the first lines of my first article on Mobile Time I asked Claude, Anthropic's Generative Artificial Intelligence (AI) assistant, to generate a haiku about artificial intelligence. The result: Synthetic reasoning / Machines gain intellect / AI, uncertain future.
We could philosophize whether there is a future that is not uncertain, but some are certainly more uncertain than others. I believe this is the case with the regulation of artificial intelligence in Brazil and its impacts on the economy.
Many of the uncertainties arise from the fact that the discussion about the need, urgency and model of a regulatory framework for AI in Brazil remained, for longer than it should have been, restricted to the interest of jurists – after all, they are the authors of the main bill under discussion today, PL 2338/2023 – and technology companies. It turns out that almost every company is a consumer of technology and, potentially, is already or will be a user of AI applications, at the very least. After all, AI will increasingly be a differentiator in productivity and competitiveness. As the topic seems distant from the core business for different sectors, education, industry 4.0 and agribusiness organizations, for example, may not have yet analyzed the impacts of AI regulation, along the lines of PL 2338/2023, on their businesses.
In 2022, when the European Union was debating its regulation for AI, a survey of 113 startups in the region revealed that 96% of them were AI providers and 65% of them were familiar with the regulatory text under discussion. For this reason, 51% of startups estimated that they would slow down the pace of AI development due to the prescriptive nature of the legislation and the costs of adapting to the standard's obligations.
In Brazil, research with the universe of startups, in 2021, he pointed out that only 4% were well aware of the AI regulation proposal being debated here in the country at that time (it was PL 21/2020, from the Chamber of Deputies). After three years, how many startups are today familiar with the risks of PL 2338/2023 and the potential compliance costs they will incur?
Another survey, carried out in 2023 with micro, small and medium-sized companies (MSMEs) from different sectors of the economy revealed that 74% of them are using AI always or often, and that 46% of them have difficulties related to the cost of the technology. What will be the impact of compliance, and how many MSMEs will have to slow down the adoption of AI and, consequently, lose competitiveness, if the current forms of PL 2338/2023 prosper?
Considering that the most recent text of the bill, presented by senator Eduardo Gomes at the end of April, has as its main characteristic dealing with obligations and restrictions – which permeates 35% of the 438 provisions of the text, according to a X-ray published by MID; and considering that the PL also imposes very robust, complex and costly governance measures for everyone who develops, supplies and operates AI systems, this should be a topic of interest and involvement for everyone.
In agribusiness, the SEBRAE estimates that AI adoption will grow by 25.5% by 2026. However, the use of AI in the field may be discouraged due to the format of regulation. Large-scale AI systems that analyze large volumes of agricultural data to predict yields or detect diseases and that can affect many people indirectly may be classified as high risk if they are not sufficiently transparent or auditable. This would require significant investments in compliance robust measures to ensure that these AI tools meet new regulatory requirements, which would increase operational costs, in addition to the risk of its agents being liable for possible damages to third parties, even if it is due to operational failures, regardless of fault or intent.
Even in the case of AI solutions whose risk is not considered high, the liability regime proposed in PL 2338/2023 imposes the presumption of guilt on the part of the agent and the reversal of the burden of proof. A disincentive for small farmers.
The best public policies and regulations are based on evidence and are built with broad debate and stakeholder involvement. Public consultation mechanisms are often used for this purpose. In the case of AI, everything indicates that the Executive Branch does not consider it necessary to expand effective listening to the sectors that will be impacted – potentially, all of them. There is a sense of urgency from Executive leaders to approve the AI regulatory framework later this year; an urgency that is incoherent and that runs at a gallop in the opposite direction of the regulatory models being developed in the countries with the greatest protagonism in AI in the world, such as the United States, United Kingdom, Japan and Singapore, among others.
This way, as my haiku written by Generative AI would say, the future will be truly uncertain.
*Cibele Perillo is Public Policy Lead at AWS and leader of the Artificial Intelligence Working Group of 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 
