Artificial intelligence (AI) is a technology widely implemented across multiple spheres of everyday and institutional life. Despite this, its adoption has taken place in an unregulated manner, without minimum safeguards to protect the population from the potential negative effects of the technology, which have already been widely identified. Strengthening studies on these impacts and proposing governance measures are essential in this context, especially given the economic power and lobbying capacity of big tech companies.

The Emerging Challenges of Artificial Intelligence project aims to identify, investigate, and analyze the main problems and risks associated with the current use of AI, with a focus on the protection of rights. Complementary objectives include scientific dissemination and the promotion of educational campaigns aimed at both specialized audiences and the general public, as well as exerting pressure on the sectors responsible for governance measures and risk mitigation.

The areas of action addressed within the project include facial recognition; the environmental cost of AI, with a focus on data centers; gender-based violence facilitated by AI tools, such as deepnudes; AI governance and regulation, with emphasis on Bill No. 2338; and copyright.

Among the main emerging challenges identified, three stand out: the existence of discriminatory biases in AI systems that can exacerbate structural social problems; the environmental impacts resulting from the infrastructure required for AI to function, especially in times of climate crisis; and epistemological impacts, which concern the reliability of disseminated knowledge, the presence of ideological biases, the spread of disinformation, and cultural erasure.

Although the entire population is affected by these challenges, the most vulnerable groups—considering criteria such as class, race, ethnicity, gender, sexuality, and regionality—are particularly sensitive to these effects. Even people without direct access to smartphones or the internet may be indirectly impacted by AI.

Emerging challenges differ from the classic problems of AI, as they reflect a scenario of accelerated development beginning in the 2020s, especially with advances in deep learning and computational infrastructure. The massive availability of data from social networks has contributed to the training of large-scale models. This has intensified already known issues such as privacy, bias, labor precarization, and environmental impacts. Today, AI is present in various everyday tools, such as mobility apps, social networks, and recommendation systems, influencing everything from social interactions to political debates.

The solutions proposed by the project are multidisciplinary in nature, encompassing technical, ethical, regulatory, and social aspects. Institutional partnerships include the Laboratory of Public Policy and Internet (LAPIN), the Brazilian Institute for Consumer Defense (Idec), and the AI Working Group of the Rights in the Network Coalition (CDR). Funding is provided by Google and the Heinrich Böll Foundation.

To date, the project has produced several relevant findings, including:

• Contributions to the Brazilian Senate’s public consultation on the development of AI regulation in Brazil, addressing privacy, bias, transparency, and civil liability.

• Texts discussing the non-neutrality of technologies, the limits of Asimov’s Laws as a basis for AI regulation, and the need to apply the principle of vulnerability to the use of AI.

• Critiques of how experts and companies influence the public debate on AI, reinforcing private interests and spreading disinformation.

• Questioning of the myth of AI exceptionalism and of the impact of deepfakes on elections.

• Technical notes on the risks of facial recognition and gender-based violence related to pornographic deepfakes.

• Analyses of the obstacle that trade secrecy represents for algorithmic transparency.

• Critical assessments of Bill No. 2338/2023, proposing amendments to balance innovation and the protection of rights.

• Reflections on the relationship between AI, human rights, and corporate responsibility, as well as suggestions to the Brazilian Data Protection Authority (ANPD) regarding automated decision-making.

• Discussions on techno-solutionism, AI explainability, the epistemicidal impacts of large language models (LLMs), and the environmental effects of data centers.

Institucional