From the introduction of the printing press to the digital age, scholarly publishers have always been at the forefront of technological innovation. They were the first publishers to go digital, and continue to incorporate new technologies, such as machine learning and artificial intelligence (AI)— increasingly including large language models (LLMs) and generative AI tools—to enhance access, quality, and integrity and to accelerate innovation. In 2021, STM published a white paper that discusses how STM publishers contribute to the ethical and trustworthy development, deployment, and application of AI. That paper addresses both the potential opportunities and risks concerning the application of AI within the scholarly ecosystem and defines best-practice principles for an ethical, trustworthy, and human-centric AI. While the principles in that paper still hold, more specific opportunities and threats have emerged in relation to generative AI (GenAI).
The emergence of GenAI has significantly increased the need to carefully consider the use of AI in scholarly publishing – with respect to both its potential and its risks. GenAI can improve the quality of published material, break language barriers, and realize speed and efficiency improvements for all stakeholders in the publishing process. It can also blur the line between machines and researchers, fact and fiction (in the form of so-called hallucinations), and allows people to fake research results and outputs on an unprecedented scale. An additional challenge is that current integrity-checking tools are not adapted to GenAI, and it may be difficult to develop that capability.
Serious questions exist regarding intellectual property rights, confidentiality, privacy, and security. For example, concerns
have been raised about whether intellectual property rights were respected in the building and training of the LLMs that
underlie GenAI tools.Litigation regarding the use of copyrighted works in training data for these tools is ongoing. Significant questions have arisen regarding how the submission of information to GenAI tools, e.g., in the form of prompts, respects laws and conventions around confidentiality and security.
In reaction to these developments, various initiatives have been launched since the public release of GenAI tools (most notably, the public release in October 2022 of ChatGPT powered by GPT3 and successive models). In August 2023, a coalition of organizations, of which STM is a part, released Global Principles of Artificial Intelligence. These principles concern the governance of the development, deployment, and regulation of AI systems and applications and cover issues related to intellectual property, transparency, accountability, quality and integrity, fairness, safety, design, and sustainable development. Individual publishers have also developed their own policies on how researchers, editors, and readers might use GenAI appropriately and ethically and are exploring additional ways that GenAI might be used in their workflows.
Developments in GenAI are highly dynamic and changing with unprecedented speed. For this reason, predicting how this technology will progress is difficult. This document reflects STM recommendations for best practices in the use of GenAI in the publication process that are based on the state of the technology at the time of its completion and, it is hoped, reflects general principles that will remain relevant in the near term. Drawing from guidelines that have been developed by individual publishers as well as current understanding of the legal, ethical, and regulatory issues around GenAI, these recommendations are intended to help individual publishers develop policies for the use of GenAI by authors, editorial teams, reviewers, vendors, third-party providers, and publishers themselves. STM expects to update these recommendations regularly because of the highly dynamic nature of these technologies.
