How is AI reshaping scientific practice—and why does this require governance and ethical reflection?

Incorporating artificial intelligence into research routines increases productivity but poses new challenges for transparency and institutional accountability

Bruno de Pierro

A man in an office works on a laptop with code and text on the screen, with books and a clock in the background. The image portrays the use of digital technologies and programming in the context of research and data analysis.

Artificial intelligence (AI) is no longer merely a subject of theoretical debate or a potential futuristic application in science. The technology now lies at the center of a significant transformation in how research is conducted, not only because it speeds up operational tasks (such as literature reviews and raw data analyses), but also because it challenges how science itself is structured and managed.

The shift presents researchers with an important question beyond merely how to use AI: who defines the limits, objectives, and rules governing its use? This is a question that encompasses technology, ethics, institutional governance, and public policy, and that was discussed in a recent Science Arena report..

AI as a driver of productivity and methodological change

Generative AI and data analytics tools are already part of everyday scientific practice. They assist with tasks that might previously have required weeks of human work, such as structuring texts, organizing reference databases, or identifying patterns in large volumes of data, helping to speed up processes and expand exploratory capacities.

However, this technological integration goes beyond efficiency alone—it is changing the methods of knowledge production.

Researchers can now use predictive models, automated information extraction, and advanced pattern analysis to formulate hypotheses and direct experiments in new ways, establishing a potential new hybrid methodology based on the interaction between humans and AI.

In fields such as bioinformatics, genomics, and big data analysis, AI is almost indispensable, said veterinarian Rebeca Scalco, a PhD candidate in digital pathology and bioinformatics at the University of Bern, Switzerland, in an interview with Science Arena.

“The AI tool SciSpace, for example, helps explain complex articles, but it is still the researcher who interprets or critically analyzes them,” she says.

Governance of scientific AI: an institutional necessity

While many educational efforts and ethical recommendations focus on individual best practices, such as declaring or recording when and how AI is used, the broader discussion revolves around governance and institutional policy.

In recent years, several universities and research centers worldwide, including Brazilian institutions such as the University of Campinas (UNICAMP) and the Federal University of Bahia (UFBA), have begun creating formal guidelines on the use of AI, with a focus on:

Transparency and accountability, requiring researchers to document and justify the role of AI in their results.
Data protection and confidentiality, preventing the exposure of sensitive information on platforms outside the institution’s oversight.
Preservation of the researcher’s authorship and intellectual agency, ensuring that AI does not compromise the originality and integrity of scientific work.

These guidelines are not just recommendations. Many universities are discussing how to incorporate AI into their internal research regulations, with specialized committees and official documents to guide students, faculty, and research groups.

AI, public policy, and global regulation

On an international level, the United Nations Educational, Scientific and Cultural Organization (UNESCO) has issued global recommendations encouraging the development of ethical frameworks for AI in science and society, emphasizing that the technology needs to be used in line with human rights, data protection, and shared ethical values.

In some regions, such as the European Union, specific legislation is already being created to define risks, transparency requirements, and oversight mechanisms.

In other countries, draft legislation addresses governance, civil liability, and ethical principles that can guide the use of AI in scientific research and technological innovation.

Challenges and conflicts: autonomy, corporate power, and scientific sovereignty

The debate about AI governance is not neutral, Scalco warns. “Academia and civil society need to be involved in regulation, ensuring that the public interest remains central, especially in fields like health,” she says.

Most advanced AI tools and platforms are controlled by private corporations (the so-called big tech companies) with global economic interests, sparking concerns about who controls the technology that is now influencing the production of scientific knowledge.

This conflict raises additional questions:

Should science relinquish sensitive, long-term data to external platforms?
What criteria should academic institutions adopt to protect their researchers and research subjects?
How can the use of AI be balanced with preservation of the critical thinking, creativity, and methodological rigor that characterize science?

The role of researchers in formulating governance

The ethical governance of AI in science is not solely the responsibility of committees or regulatory bodies.

Researchers should play an active role in developing standards, policies, and practices that reflect the values of the scientific community: transparency, accountability, equity, and commitment to knowledge as a public asset.

“AI tools should be treated like ‘good interns,’ eager to learn. But just as a doctoral student would never submit an intern’s unreviewed text for publication, nor should they submit work done with AI without proper verification,” says Scalco. In this sense, AI should be seen not merely as a tool, but as an agent of structural change, the ethical and political implications of which need to be understood, debated, and incorporated into institutional systems and scientific training.

* This article may be republished online under the CC-BY-NC-ND Creative Commons license.
The text must not be edited and the author(s) and source (Science Arena) must be credited.