Editorial CSIC, following the recommendations of the Committee on Publication Ethics (COPE) as its main reference, understands that the use of artificial intelligence (AI) and AI-assisted tools can serve as auxiliary technology, a useful support, provided that they are used responsibly and prudently, their use is disclosed in detail, and their outputs are reviewed, corrected, and accepted by human beings. This duty of transparency is also in line with the principles of open science, which promote the traceability and reproducibility of research.

The ethical and responsible use of these technologies requires constant oversight and supervision by those responsible for authorship, editing or reviewing. These individuals must verify the accuracy, precision, validity, relevance, and integrity of the material generated by AI, paying special attention (though not exclusively) to fabricated references and biases that may arise from the tool’s design (such as inconsistencies, lack of information, or missing sources) to prevent misinformation. Accountability in the use of AI tools must ensure thorough analysis and disclose any detected limitations.

The most common types of AI in the field of scientific publishing are generative AI and Large Language Models. However, many other types of AI can be used in the processes of manuscript writing, editing, and production, and even earlier, during the design or development of the research.

There is a wide variety of tools and technologies that use AI for editing, and many of them allow for the protection of the content we input. It is essential to use tools that enable encryption of communication or stored information, deny authorization for the model to be trained on our content, or operate in controlled environments, whether on a local installation or on servers with privacy guarantees, as well as any other protective measures.

For a detailed description of Editorial CSIC’s policy on the use of artificial intelligence, see the Good practice guide for scientific-academic publishing, which describes the concepts of authorship in relation to AI, who should declare the use of AI and how it should be declared, and how AI tools should be referenced and cited.