Brazil Case

REGULATION

The Brazilian AI regulatory framework is currently evolving, with several key bills under discussion. However, as of September 2024, there was no fully approved national regulation in place. The only AI-related regulation that has been enacted is at the subnational level, in the state of Ceará. At the national level, there are 15 proposed bills related to AI in the Federal Senate (Higher Chamber) and nine in the Chamber of Deputies (Lower Chamber). This legislative activity is grounded in the General Data Protection Law (LGPD), which has been in effect since August 16, 2020. The LGPD was created to unify and modernize the 40 previously existing laws concerning the processing of personal data, aligning them with the realities of an increasingly digital society. Modelled after the European GDPR, the LGPD establishes principles such as respect for privacy, freedom of expression, and the protection of human rights, which guide the ongoing debates on AI regulation in Brazil.

There are isolated regulatory initiatives in Brazil, the most complete in content and advanced in terms of implementation being Bill 21/2020, proposed in the Chamber of Deputies on February 3, 2020, and approved by this legislative house on September 29, 2021. This bill was later attached to Bill 2338/2023, which is now the main focus of discussions at the Temporary Inner Commission on Artificial Intelligence in Brazil, created by Brazilian senators to discuss and propose a national AI regulation law. The expectation is that, once federal legislation is fully approved, it will serve as a regulatory framework for the sector, addressing key issues such as the protection of fundamental rights and the correction of abusive practices. .

The Bill 2338/2023, also known as the “AI Bill”, represents a significant step in regulating artificial intelligence within the country. The intention is to consolidate it into new legislation that encompasses many of the previous proposals. Originating from Brazilian Strategy of AI (EBIA), the bill outlines general principles, including the relationship between AI and work, though it remains broad and lacks detailed regulations. The Federal Senate established a legal experts commission to refine the bill, focusing on ensuring that various sectors, such as healthcare and telecommunications, develop specific AI regulations. This process is ongoing, with further public and international input being sought, though there has been little discussion on labour rights within the AI context.

The AI Bill comprises 16 articles which establish the scope of the law, define technical terms and outline principles, rights and obligations related to AI. While the initial articles cover fundamental aspects broadly, later sections provide more operational details, such as aligning with the General Data Protection Law (LGPD) and setting guidelines for the role of government entities in AI development and regulation. The bill emphasizes investment in AI research, restructuring the labour market, and promoting transparency and ethical practices in AI use by public bodies. Additionally, it highlights the State's constitutional duty to provide education, particularly in training individuals for the responsible use of AI, underscoring the importance of capacity-building in this rapidly evolving field. As of August 2024, the bill was receiving amendment proposals –  145 so far – on various subjects, including the protection of workers from algorithmic decisions, the use of AI in elections, and many others, which are still to be debated within the Federal Senate.

At the same time, other more specific bills were proposed recently and are being discussed both in the Chamber of Deputies and in the Federal Senate. Some examples are the Bill 266/2024, from the Federal Senate, that aims to provide legislation for the use of artificial intelligence systems to assist the work of medical doctors, lawyers and judges; the Bill 536/2024, from the Chamber of Deputies, which proposes regulations for the profession of “Autonomous Driver of Urban Mobility Services”; and the Bill 4869/2023, also from the Chamber of Deputies, which prohibits the dismissal or reduction of working hours of voice professionals and Brazilian Sign Language interpreters due to the use of artificial intelligence tools to produce audio-visual content. The two latter are being processed in the Chamber of Deputies.

Beyond the bills proposed directly at the federal level, there are also bills proposed in Brazil’s federal unities’ legislative assemblies aimed at regulating AI at state levels. There were four initiatives in this regard: in the states of Rio de Janeiro, Minas Gerais and Ceará, and in the Federal District. As mentioned before, Ceará is the first and only state in Brazil to have enacted legislation specifically addressing artificial intelligence, through Law 17611, dated August 11, 2021. The law, consisting of four articles, emphasizes the need for AI systems to be designed ethically, safely, and in compliance with Brazilian laws. It defines what constitutes artificial intelligence under its jurisdiction and provides operational guidelines for companies developing AI systems.

Key directives stipulated by the law include ensuring the security and privacy of user data, allowing users control over their personal information, and maintaining human oversight over AI systems. The law also mandates that AI systems respect human dignity, avoid discrimination, and promote social well-being, while not inciting hatred or violence. Additionally, it requires companies operating in Ceará to be held accountable for any damages caused by their AI systems. However, the law does not stipulate the need for a strategy or mechanism to previously detect these system failures, leaving a potential gap in its regulatory framework.

The development process of AI legislation in Brazil reflects the country’s commitment to addressing the challenges and opportunities presented by this rapidly evolving technology. As discussions continue at both federal and state levels, Brazil is working towards a comprehensive regulatory framework that balances innovation with the protection of fundamental rights. The various bills under consideration, including Bill 2338/2023, highlight the need for detailed regulations across different sectors. As amendments are debated and local initiatives like Ceará's pioneering law take effect, Brazil is attempting to lay the groundwork for responsible AI governance that can adapt to future advancements. 

DEVELOPMENT

In recent years, Brazil’s AI field has developed reasonably. In terms of patents, 550 applications related to AI were filed in the country in 2019 against 150 per year in 2010. Notwithstanding, this number is still low compared to Europe’s (10,500) and the United States’ (43,000) in that same year. Moreover, of the 20 companies that submitted the most national AI patents’ requests, more than half were from the United States (12), with the rest being from Europe (3), Japan (3), India (1), and China (1). This means Brazilian companies are not anywhere close to leading innovation in AI.

As for AI funding, the numbers show a significant rise. This was accompanied by a growing adoption of AI systems throughout the business environment, albeit at a slower pace than the Global North. In 2023, 13% of Brazilian companies reported adopting these technologies. However, this rate was considerably bigger in the case of large companies (250 people or more): 41%. The sectors with the highest rates of adoption of these technologies were information and communication (40%); professional, real estate, scientific and technical activities (18%) and transportation, storage and postal services (14%).

According to a 2022 survey, a total of 702 companies working with and developing AI were registered in Brazil in that year. Out of these, 223 companies were multisectoral and 479 focused on specific sectors. The main sectors were AI as a Service (76%), Business Intelligence and Analytics (69%), Health and Biotech (60%), human resources (48%) and Industry 4.0 (46%).

Another survey mapped 163 Brazilian companies with a focus on AI development. The work also showed that Brazil was the Latin American leader on this matter. Chile and Mexico, respectively the 2nd and 3rd, had 42 and 36 AI companies. Despite flaunting the regional leadership, Brazil was well behind China and the United States, which had, respectively, 861 and 5,068 AI companies. The advancements on the AI field, nationally and internationally, triggered a series of AI-related policies in the country. One good example are the 11 Applied Research Centres in Artificial Intelligence fostered by the Federal and São Paulo State governments in partnership with private companies. Another important initiative is the IA² MCTI (Ministry of Science, Technology and Innovation) program, which has supported 127 AI start-ups since 2022. 

It is also worth mentioning the Brazilian national strategies related to AI development. The first one is the Brazilian Strategy of Digital Transformation (2018), which lists AI as a key technology for development among others. The second one is the Brazilian Strategy for Artificial Intelligence (2021), which establishes general guidelines for AI development in Brazil.

In July 2024, the Brazilian government launched its main AI strategy, the Brazilian Plan for Artificial Intelligence (PBIA). Encompassing the period of 2024-2028, the plan is based on five general goals:

1. transform the lives of Brazilians through sustainable and inclusive innovations based on AI;
2. equip Brazil with advanced technological infrastructure with high processing capacity powered by renewable energies;
3. develop advanced language models in Portuguese, with national data that considers cultural, social, and linguistic diversity, aiming at strengthening AI sovereignty;
4. train and requalify people in AI on a large scale to value the workers and meet the high demand for qualified professionals;
5. promote Brazil's global role in AI through national technological development and strategic international collaboration actions.

The general goals should be met by planned actions, which are structured in five different axes. They are:

1. infrastructure and development;
2. dissemination, training and capacity building;
3. improvement of public services;
4. business innovation; and
5. support for regulation and governance.

Some of the initiatives projected entail the creation of AI infrastructure programs, the structuring of the data and AI software ecosystem, the purchase of a supercomputer to boost cutting-edge AI research, the development of high-performance processors and the constitution of a national network of AI centres of excellence. The plan also envisages policies that back start-ups, support national data centres and promote AI systems for micro and small-sized companies.

Investments of R$ 23 billion (USD 4.25 billion) are planned, of which R$ 13.9 billion (USD 2.51 billion) should go for business innovation, R$ 5.8 billion (USD 1.04 billion) for infrastructure and development, and R$ 1.15 billion (USD 200 millions) for training human resources. More than 50% of the resources shall come from public funding and institutions (e.g., the National Economic and Social Development Bank and the Studies and Projects Funding Agency), and almost 35% of the public funding shall come from the National Fund for Science and Technology and from the annual federal budget. Private industry shall contribute with around 4% of all resources.

The world of work is barely addressed in PBIA. When mentioned, it is related to training, qualification and requalification of workers to address the high demand for AI professionals. The most prominent planned initiative on the topic is a platform to connect job vacancies with candidates registered in the national employment system. The plan also vaguely addresses expressions such as “valuing workers” and “national sovereignty”. 

In Brazil, using the "regulatory sandboxes" instrument to make legal and regulatory obligations more flexible has also gained strength in recent years. In 2023, the National Data Protection Authority (ANPD) launched and put out for public consultation a regulatory sandbox program on Artificial Intelligence. The entity listed benefits for organizations and companies, such as: Contributing to innovation and the development of products and services and identifying good data governance practices, among others. On the other hand, the Authority would have the following gains: Learning about innovations first-hand and promoting projects that respect data protection. When this report was published, the ANPD had not officially launched the program.

Regulatory sandboxes for AI are also part of the AI bill being debated in the Federal Senate (Bill 2338/ 2023). According to the bill, the competent authorities must set the rules for the adoption of regulatory sandboxes, providing priority access to micro and small-sized companies. The bill also determines that governments will adopt measures and foster the development and innovation of AI systems in the productive sectors. Among the planned actions are the financing of development projects and incentives for the adoption of computer infrastructure.

However, despite the measures to support the private sector in the bill, in 2024 it was not voted due to the lobby of the Big Tech companies and the National Industry Confederation (CNI) and the pressure on senators. The National Industry Confederation argues that the bill imposes too many regulatory obligations, which would make the country's productive sector "lose competitiveness and opportunities".

GOVERNANCE

AI governance is not defined only by States and companies. Social movements, labour unions, business associations, NGOs and think tanks also seek to dispute the meanings of AI governance, through different forms of action, advocacy and bargaining. Thus, different stakeholders try to push for AI governance that is fairer according to their interests.

In the case of Brazil, there is a solid tradition of civil society advocacy on issues related to digital rights. The Brazilian internet governance, which consists of a pioneering multisectorial composition, is a good example of this. As for AI governance, several organizations, of different natures, have been authoring public and collective letters expressing their stance on the matter towards policy makers and governments. The analysis of some of these letters suggests there has been no agreement between labor unions and business associations regarding the impacts of AI on the world of work.

The main reason for the mobilization of unions and worker associations was the beginning of discussions regarding AI regulation. The Central Unified Workers' Union (Central Única dos Trabalhadores, CUT), the main Brazilian trade union, published a letter at the end of 2023 fully supporting Bill 2338/23. The letter warns of the dangers of AI-powered fake news and the misuse of databases, but states that the bill’s articles provide adequate regulatory risk assessment and algorithmic impact mitigations. It also highlights the notion of global cooperation and advocates that AI solutions shall be negotiated with workers' representatives.

There was one specific sector that raised concerns about the direction of AI regulation: the cultural sector, especially the film, audiovisual and music industries. This is possibly a consequence of the impact of the 2023 strikes of the Hollywood workers - especially writers and actors - regarding the use of generative AI at the workplace. In April 2024, the Union of Workers in the Film and Audiovisual Industry of seven Brazilian states published a letter addressed to the Federal Senate’s president highlighting the lack of discussions around AI impacts on the Brazilian audio-visual industry, considered by them to be one of the main drivers of the country's economy and culture.

In another letter, also published in April 2024, 26 professional associations from areas such as journalism and music demanded that Bill 2338/23 should protect rights from their workers’ segments. Among other topics, the letter advocates the need for consent of copyright holders for the use of their work for AI training in addition to fair remuneration and transparency in relation to the use of AI. In July 2024, 24 other associations, mainly of cultural workers, published a similar letter expressing concern about copyright and the preservation of intellectual property by creators. The main issues highlighted was remuneration and awareness over the use of creators’ works for data training by tech developers.

Voice actors also got together on the Dublagem Viva movement in order to address the impact of generative AI on their work. They published a manifesto, signed by several organizations and supported by United Voice Artists, focusing on the need to not replace workers with AI. They also claim that generative AI should be understood within the Brazilian legal and cultural frameworks, linking generative AI to a possible loss of national and cultural identity.

Business associations have also expressed their views on the country's AI governance process. In general, their positions state that the regulatory proposals are among the most restrictive in the world and may inhibit AI development and innovation in Brazil. But they also requested the participation of as many actors and institutions as possible – including academia and the technical sector – to decide on the direction of AI development in the country.

In 2023, the Brazilian Association of Software Companies published a letter stating that the sector had not been heard by policy makers and governments. For software companies, in terms of AI, Brazil should focus on the following points: incentives for the development of AI, legal certainty, regulation based on the use of AI and not on the technology itself, regulation based on risk, allocation of duties and civil liability. The letter also highlights the importance of effective governance and coordination by Brazilian authorities in order to provide an adequate regulatory environment.

As of April 2024, 40 business associations demanded in an open letter an AI regulation that respects Brazil’s regulatory history, protects fundamental rights while ensuring the country’s socioeconomic development, innovation and competitiveness. For these associations, AI regulation should require an integrative approach among different regulatory agencies and acknowledge possible multisectoral impacts. The letter also points out the risks of dependency on foreign models, increased inequalities and the outflow of skillful AI professionals to other countries. Additionally, it stresses the necessity of investing in workforce training and of backing small and medium-sized companies.

At least an important and influential Brazilian business association was clearly against Bill 2338/23. In July 2024, the National Industry Confederation stated that regulation would hinder AI development in Brazil. On the other hand, the Brazilian Federation of Banks, one of the country’s most powerful business associations, demonstrated support for AI regulation through one of its directors. Thus, there are nuances and differences depending on the association representatives. The National Conference of Industry stated that regulation will hinder the development of AI in Brazil: "Brazil has competitive advantages and artificial intelligence has great potential to boost productivity in industry. But the proposed regulation ignores the different uses and risks and imposes a heavy regulatory apparatus not only on AI applications, but also on research and development of the technology." On the other hand, the Brazilian Federation of Banks stated that AI has been used in the financial sector for a decade and that regulation will not hinder the banks' business. This federation supports AI regulation in Brazil in the way that is being addressed by policy makers.

On the part of social movements, there is mobilization by movements such as the Homeless Worker Movement (MTST) in favour of digital sovereignty based on a popular framework (popular sovereignty), but without specifically addressing artificial intelligence. On the NGO’s civic space, the Coalition for Rights on the Internet (Coalizão Direitos na Rede) led a group of 55 other NGOs, political groups and other institutions into signing an open letter demanding a regulation on AI systems that envisages innovation while respecting human rights and transparency, and guaranteeing inclusive economic development. Published in July 2024, the letter points out that Bill 2338/23’s articles meet most of their demands, but suggests some additions to the bill’s draft in order to address issues such as facial recognition and credit scoring. Moreover, the letter demands greater participation of civil society in AI governance and oversight systems.

Finally, researchers’ associations took on an initiative calling for data workers to be recognized on Brazilian AI regulation. In July 2024, a letter was sent to senators requesting that all those who annotate data for AI applications should be included in risk assessments and in human oversight obligations for decisions made by AI systems. The letter also requested the inclusion of transparency obligations for companies regarding data workers in value chains to allow oversight by labour authorities. In addition, it demanded that data workers should have labour rights as provided for in Brazilian legislation.

CONCLUSIONS

The previous evidence demonstrates how different political and economic interests occur in discussions on AI policies in Brazil in terms of regulation, development and governance. As it stands, Bill 2338/2023 has elements that address important demands from Brazilian society. It establishes definitions of technical terms and principles, rights and obligations related to AI; emphasizes investment in AI research; promotes transparency and ethical practices in the use of AI by public bodies; and highlights the constitutional duty of the State to provide education, particularly in training individuals for the responsible use of AI.

However, the bill does not address important labour issues such as the workers’ rights in the AI value chain and the use of AI at the workplace. In August 2024, Brazil’s Ministry of Labor established an internal working group focused on AI and work. Its first meeting was attended by researchers from the AI Policy Observatory for the World of Work, who spoke about the challenges of addressing AI policies for workers considering the complexity of this topic. This working group promises to present a proposal for action by the end of 2024.

At the government level, these challenges mean ensuring clear political guidelines for AI development, for investments and for the civil and labour rights of the population. Initiatives in this area date back to 2018, but were revised in 2024 and are disclosed in the Brazilian Plan for Artificial Intelligence, which was described in the Development section:

1. infrastructure and development;
2. dissemination, training and capacity building;
3. improvement of public services;
4. business innovation;
5. support for regulation and governance.

In the infrastructure and development dimension, there is a plan for the creation of infrastructure programs for AI, structuring the data and software ecosystem for AI, and AI research and development. In the industrial innovation dimension, the government promises to launch programs to promote the AI value chain and develop systems for the challenges of Brazilian industry and increase productivity.

If the plan’s guidelines are adopted and implemented, Brazil should advance even more its AI leadership in Latin America in relation to the AI value chain. It should also be able to advance in the levels of its competitive advantages and act with greater sovereignty in relation to the interests of hegemonic countries and regional blocs such as the USA, China and the European Union. Last, but not least, it should have the ability to position itself in a more independent and autonomous way in technological development and innovation in relation to large tech conglomerates, who nowadays control knowledge and wealth in the AI value chain.

The issue of AI governance is also complex. The term governance itself may not capture exactly how interests in AI are embedded in the class struggle itself – or how negotiations, advocacy, lobbying and bargaining are embedded in power relations. We have expanded the definition of this term to address how different civil society entities – through social movements, unions and associations –  have been working together  to express their voice and participate in AI governance. The demands that might arise from these stakeholders are numerous and diverse, such as the promotion of an AI literacy framed into a popular perspective in order to fully understand the AI value chain; the participation of workers’ and social movements’ representatives in decision-making processes regarding AI and economic development to ensure all perspectives are considered, such as the gender and racial ones; and the regulation of social media and digital labour platforms in order to guarantee workers’ rights, which also embraces the data they produce.

In Brazil, civil society actions are still timid and are more institutionally positioned in dialogues with the government and National Congress. There is an intense movement of workers fighting for their rights in relation to AI, but this is still in an early and learning phase in terms of its organizational forms, which might take shape as strikes and digital mobilization. Other types of mobilizing and organizing are still embryonic and experimental, such as the promotion of an open debate regarding workers’ cooperatives in the AI value chain, and need greater synergy with experiences from other regions. Nevertheless, initiatives such as the one featured by the cultural workers can be important to establish a framework of demands that aims at protecting workers’ rights in the context of AI development.

To sum up, the current AI regulation under discussion in Brazil, framed at Bill 2338/23, still does not adequately address the world of work concerning either labour rights and data rights. This leads to the inevitable conclusion that important challenges remain to be tackled in the short term, both by local and global agents, in order to guarantee stronger AI policies for the world of work in Brazil.

References

• ABStartups (2022). The impact and future of Artificial Intelligence in Brazil. ABStartups and GoogleforStartups. Available on Google Drive.
• ANPD (2023a). Regulatory sandbox and data protection in Brazil. National Data Protection Authority (.PDF).
• ANPD (2023b). ANPD extends consultation to society on Regulatory Sandbox until December 1st. National Data Protection Authority.
• Brandão, Rodrigo (2024). The current scenario of development of Artificial Intelligence in Brazil. Sectoral Overview of the Internet, Cetic.
• Cetic (2023). ICT Companies 2023. Internet Steering Committee in Brazil.
• CNI (2024). Artificial Intelligence Bill harms the development of technology in Brazil and innovation in the productive sector.
• Machado, T. M., & Winter, E. (2023). Artificial Intelligence and patents in Brazil: Overview on patentability and comparative study on patent filings. World Patent Information, 72.
• Mello, Patrícia Campos (2024, 9 jul). Vote on AI project is postponed in the Senate for the fifth time, after pressure from CNI. Folha de S. Paulo.
• Ministry of Science, Technology and Innovation. (2018, 21 Mar.). MCTIC Ordinance No. 1,556, of 03.21.2018: Approves the Brazilian Strategy for Digital Transformation (E-Digital).
• Ministry of Science, Technology and Innovation. (2021, Jul). Brazilian Strategy for Artificial Intelligence – EBIA (.PDF). 
• Ministry of Science, Technology and Innovation. (2024, 29 Jul.). AI for the Good of All Proposal for a Brazilian Artificial Intelligence Plan 2024-2028.
• Ministry of Science, Technology and Innovation. (2021). Brazilian Strategy for Artificial Intelligence (.PDF)
• Digital Development Observatory. (2024, 9 feb). Startups drive AI-based innovation in the region.
• OECD (2024). OECD. AI Policy Observatory: Brazil. Retrieved: 20 August 2024.