It is not our aim, in this editorial, to offer a final definition of research ethics or ethics in scientific publishing, especially at a time when we are increasingly connected and are still learning to deal with the rapid growth of artificial intelligence (AI). It would be unreasonable to assume that readers are unfamiliar with the stages of research or with the ethical principles that should guide scientific work. Researchers can usually explain what they study, which methods they use, and how their collaborations are organized. They also know, at least in general terms, which rules must be followed. Yet knowing these elements does not remove the questions that arise during research or, by itself, ensure that every decision will be right.

What changes, then, when AI becomes part of this process? Perhaps the more useful question is not how to place ethics and AI on opposite sides, as if we had to choose
between them, but what AI changes, intensifies, or makes more urgent in scientific practice.

A scientific research can be understood as a form of “intellectual craftsmanship”:  patient and creative work in which theory and practice are closely linked and require discipline, methodological rigor, and critical judgment.^1,2 Researchers do not simply apply a technique to a readymade object. They turn questions, observations, and concerns into an ongoing process of investigation.

Although this idea comes from the humanities and social sciences, it applies across different fields of knowledge. In any field, method and theory depend on each other. Without theory, technique loses direction; without method, interpretation loses support. Scientific work also requires a clear record of the process: carefully documenting what was done, returning to earlier ideas, recognizing changes in direction, and maintaining a dialogue with other researchers.^1,2

If we have become scientists, this is the result of our paths, the opportunities we found, and the choices we made – or that circumstances allowed us to make. We do not stand outside history, and our training and surroundings inevitably shape us. This does not mean, however, that data may be adjusted to fit our beliefs. It means that our choices must be examined and our interpretations must remain open to criticism.

Science has always relied on instruments, laboratories, databases, computer  programs, technical staff, collaborators, reviewers, publishers, and institutions. These elements do not automatically make a conclusion correct, but they take part in the production, evaluation, and circulation of knowledge. The history of chemistry  itself shows this collective nature: the quantum model we teach today was not the work of one person, but the result of contributions built up over time. Scientific knowledge can therefore become part of humanity’s shared heritage.

Science is a human activity and is therefore fallible. Recognizing this does not weaken science. On the contrary, it supports systematic doubt, checking by others, and the possibility of self-correction. These are our main safeguards against what we might call “hasty certainty.”^3,4

In Max Weber’s account, the ethics of responsibility offers a useful way to think about scientific practice.⁵ Good intentions are not enough: we must also consider the foreseeable consequences of our choices and answer for them. Scientific integrity requires respect for facts, consistency between evidence and conclusions, and honesty about what the data do – or do not – allow us to claim. Values are present when we decide which questions deserve attention, which risks are acceptable, how credit is shared, and which benefits or harms may result from research. They must not, however, determine in advance the result we hope to find.

Ethical commitment comes before any code of conduct, committee decision, or editorial guideline. This does not make such rules unnecessary. Personal beliefs and common sense, although important, are not enough to sustain public trust in science. Research and scientific publishing need shared rules, clear procedures, and a clear assignment of responsibility for each decision.^3,4

This is where artificial intelligence becomes especially relevant. AI did not bring technology into science; technology has always been part of it. What AI has changed is the speed and scale of many scientific activities. Tasks once completed through several steps can now be automated, combined, and repeated in seconds. In a system in which professional recognition is closely tied to producing and publishing results, it is understandable that these tools have been welcomed with enthusiasm.

Rejecting every AI tool simply because it is a tool would be like refusing to use a ruler because it can be misused. A ruler can help us draw a line, but it does not decide where that line should begin, where it should go, or whether drawing it makes sense. Something similar happens with AI. The ethical issue lies not only in whether the tool exists or is used, but also in its purpose, the data it receives, the conditions under which it operates, what is delegated to it, and how its outputs are checked.^6,7

In a highly connected environment, errors and biases can move quickly among articles, databases, review reports, and the media, often before they are identified. When we give up the careful, “craft-like” work of checking what AI produces, we risk repeating information that only appears to be correct. The main question, then, is not simply whether AI was used. We must ask what was delegated, why it was delegated, which tool was used, whether its use was disclosed, who checked the output, and who will take responsibility for it.⁶ In other words, the challenge is acceleration without abdication.

These questions become clearer when we follow the course of a research project. Ethical commitment begins at the planning stage, with clear objectives and a suitable method. If a study involves humans or animals, it must follow the applicable rules, protocols approved by the relevant ethics committees, and, when required, formal consent procedures.^3,4

The collection, recording, and analysis of data also require care and complete  honesty. Data must never be adjusted to produce the desired finding. Unexpected or
inconsistent results must be investigated, not hidden. It may be necessary to repeat measurements, check instrument calibration, review experimental procedures, and, in chemistry, purify and characterize products again. Criteria for excluding data must be defined and justified, and negative results should not be discarded simply because they do not support the original hypothesis.^3,4

The aim is not to obtain the expected result, but to produce reliable and traceable data that are, whenever possible, reproducible. Honest errors can occur and should not automatically be treated as misconduct. Science depends on its ability to correct its own errors. The ethical problem begins when information is hidden or when data are deliberately neglected, fabricated, falsified, or manipulated.^3,4

Research ethics and publication ethics are closely linked, but they are not exactly the same. Research ethics applies to the question being asked, the design of the study, and the collection and interpretation of data. Publication ethics adds duties related to the scientific record, authorship, transparency, peer review, and editorial decisions. An ethical publication depends on ethical research, but it also requires its own safeguards.^3,4,8

This commitment continues at the time of publication. We publish research not only to gain academic recognition, but also to contribute to a field of knowledge. Other researchers will have access to the work, may use it as a reference, and will often try to reproduce it. When data are altered or have not been checked with proper care, the consequences may include corrections, retractions, and loss of trust in the authors, their institutions, and the scientific literature itself.^3,4,8

Authorship is part of this commitment. Criteria may vary across fields, but the basic principle is simple: the author list should reflect real scientific contributions. In many fields, the first author is the person who carried out most of the research, performed the experiments, interpreted the data, and prepared the first draft of the manuscript. The other authors should have made substantial contributions to the design, execution, or analysis of the study, or to writing and critically revising the work. All authors must approve the final version and be willing to take responsibility for the published content.^3,9

A large number of authors is not, by itself, an ethical problem. Some studies truly require large and diverse teams. The problem arises when authorship is given as a favor, a sign of status, or a form of repayment, without a real contribution. It is also wrong to exclude someone who made a substantial contribution. Giving credit fairly is part of research integrity.^3,9

The same care applies to peer review. It should be a careful assessment by people with enough expertise to examine the work and help improve it. Creating fake email
addresses for reviewers, suggesting very close colleagues without disclosing the relationship, and other forms of manipulation harm the independence of the review process.^3,10

AI adds a new layer to this problem. The scientific judgment required in peer review cannot be handed over to a machine. Policies on limited support uses, such as language editing, may vary among publishers, but a manuscript under review is a confidential document and should not be uploaded to external systems that may  store, reuse, or learn from its content. Reviewers remain responsible for reading the work, checking its arguments, and explaining the basis for every  recommendation.^10-12

Conflicts of interest must also be disclosed. Their existence does not automatically mean that dishonesty has occurred. Research funded by a company that may benefit from its results, for example, is not necessarily invalid. However, the source of funding and any role played by the funder in the study design, data analysis, or decision to publish must be reported. The ethical problem lies in hiding the conflict or allowing it to influence the work without oversight and transparency.^3,4

When these principles fail, the effects go beyond a single article. Loss of trust in  science encourages science denial and makes evidence-based public decisions more difficult. During the pandemic, we saw this both in distrust of vaccines that had been shown to work and in support for medicines with no scientific evidence. Public trust does not depend on presenting science as infallible. It depends on showing that science has procedures for recognizing uncertainty, correcting errors, and holding people responsible for misconduct.^4,8

Practices that undermine scientific integrity include data fabrication and falsification, plagiarism, improper authorship, undisclosed conflicts of interest, manipulation of peer review, and irresponsible use of AI. Fabrication means creating data or results that were never obtained. Falsification means changing materials, procedures, or data to support a conclusion. Plagiarism is the use of another person’s words, ideas, images, or results without proper credit.^3,4,8

There is no universal number of consecutive words that, by itself, defines plagiarism. The assessment depends on the context, the originality of the passage, how it was used, and whether the source was properly credited. Plagiarism may involve copying online content without citation or taking ideas, images, or phrases from other  authors. Reusing substantial parts of one’s own work without disclosure and assigning authorship improperly are also integrity concerns, although they are not the same as plagiarism. In addition to academic consequences, the improper use of someone else’s work may violate copyright and, depending on the case, lead to civil or criminal consequences.^3,4,8