Upholding Scientific Rigor in the Age of Generative AI : What FASEB 's New Policy Means for The FASEB Journal and FASEB BioAdvances

The rapid rise of generative artificial intelligence (Gen AI) presents a paradox for the scientific community: a technology powerful enough to accelerate discovery, yet disruptive enough to threaten the very foundations of research integrity. The Federation of American Societies for Experimental Biology (FASEB) has taken a leadership role in Gen AI since the tool became mainstream in 2023. First to be impacted were the journals, and the FASEB Journal released our first publication guidance on Gen AI shortly after the launch of the new generation of LLMs in late 2023. The federation developed and implemented a staff policy allowing use with proper considerations and credit implemented in January 2024, and the FASEB Board of Directors appointed a 2024 taskforce that developed and published public-facing broad Recommendations for Gen AI in the Biological and Biomedical Sciences 1 that incorporate sections on Policy and Regulation, impactful topics including Scientific Integrity and Intellectual Property, Data Privacy, and other key issues. The federation also developed a separate set of society-focused guidance tailored to the 20 federation member societies. Since our first efforts, the federation has steadfastly held that we could not/should not ignore Gen AI, but we must find paths forward for the responsible use and applications of Gen AI in the biological and biomedical sciences. In 2025, a survey published by Wiley found that 73% of researchers believe the most important role for scholarly publishers to play is to provide clear guidelines on what uses of AI are acceptable in the context of publishing scholarly research 2. In response, the FASEB has introduced a comprehensive policy governing the use of Gen AI in its publications, which is derived from overarching guidance provided by the publisher Wiley 3. The policy is a statement about the values that must guide science in an era of unprecedented computational capability. At its core, the policy recognizes a simple truth. Gen AI tools can be extraordinary engines of productivity and creativity. They also introduce ethical, legal, and epistemic risks that cannot be ignored. These tools can inadvertently expose confidential information, generate fabricated or biased content, or obscure the provenance of ideas. They also raise questions about copyright, data ownership, and reproducibility. FASEB's position is that these risks are manageable, but only when authors remain firmly in control of the scientific process. One of the policy's strongest and most necessary assertions is that authors retain full responsibility for the accuracy, originality, and integrity of their work, regardless of whether Gen AI tools are used. This is not a symbolic statement. It is a reminder that scientific authorship is grounded in accountability, judgment, and expertise, all qualities that no algorithm can replicate. A recent survey of 5000 researchers published in Nature supports the integration of Gen AI to support researchers in writing manuscripts 4. Researchers were supportive of the use of Gen AI in editing a manuscript, with 70% of surveyed researchers stating this is appropriate. However, other uses of Gen AI in manuscript preparation were viewed as less acceptable, with 47% stating it is appropriate to make summaries of other papers to use in your own paper and only 36% stating it is appropriate to use Gen AI to translate your paper into a different language. Importantly, the survey reports that the use of Gen AI in any manner to prepare a manuscript should be accompanied by disclosure. FASEB believes disclosure for authors should be easy (Materials and Methods, Acknowledgments, or Figure legends) and not stigmatized. The FASEB Gen AI policy states that Gen AI can assist with drafting, editing, translating, or reorganizing text. As of this writing, it cannot evaluate the validity of a claim, verify a citation, or interpret experimental results. AI tools cannot yet understand the nuances of a research question or the implications of a methodological choice and cannot assume responsibility for errors, omissions, or ethical breaches. For these reasons, FASEB makes clear that Gen AI tools are not collaborators and not credited as authors, which complies with the position of the Committee on Publication Ethics (COPE) on Authorship and AI tools 5. Authorship carries obligations that rest with human contributors, who remain accountable for verifying claims, validating sources, complying with ethical and regulatory standards, and securing appropriate rights and permissions. Our policy affirms that scientific reasoning and responsibility remain human endeavors. The FASEB policy distinguishes between substantive use of Gen AI and routine features embedded in standard software. Spell-checkers, grammar suggestions, and predictive text do not require disclosure. FASEB journals permit the use of Gen AI tools, including large-scale language models, for tasks such as drafting, editing, translation, reorganization of text, and generation of illustrative figures. When Gen AI tools are used in manuscript preparation, authors should provide explicit documentation detailing that use. This documentation should include the tool name, vendor, version (when applicable), date accessed, and a concise description of how the tool was used in preparing the manuscript. Depending on the nature of the contribution, this disclosure should appear in the Materials and Methods section but may also appear in the Acknowledgments and/or relevant Figure Legends when appropriate. Any use of Gen AI that shapes the scientific conclusions reached in the narrative, figures that are meant to represent scientific results or findings, or data presentation is required to be disclosed. This approach strikes a thoughtful balance between information that is necessary for reviewers, editors, and readers to understand and/or repeat the science and methodology, with less of a burden on the authors to track more routine use cases that do not impact the science. The FASEB policy seeks disclosures of more substantive Gen AI use that include the tool name, vendor, version (when applicable), date accessed, and a description of how the tool was used. A central pillar of FASEB's Gen AI policy is a standardized approach to citing the use of AI in manuscript preparation. Transparency about how generative AI tools contribute to scholarly work is essential not only for reader trust but also for reproducibility, accountability, and appropriate attribution of intellectual labor. Importantly, undisclosed AI use constitutes a publication-ethics breach. FASEB will monitor this dynamic space and react accordingly, focusing on whether the results and conclusions are compromised by non-disclosure. FASEB editorial decisions may be expected to be grounded in the developing recommendations from organizations such as COPE, stating that retraction should be considered if the editors do not have confidence in the results and conclusions of the paper, including undisclosed involvement with AI 6. Providing a clear Gen AI disclosure serves several purposes. It informs readers and reviewers about the provenance of the manuscript, reinforces the author's ownership of the intellectual work, and models responsible AI use for the broader scientific community. Well-crafted disclosure statements demonstrate that Gen AI can function as a legitimate research assistant by enhancing efficiency and clarity while preserving the foundational principles of human judgment and scientific rigor. Perhaps the most consequential part of the policy concerns images, data, and figures. The new policy states that AI-generated, modified, or enhanced images may not be used as data. This prohibition is vital in that scientific images, whether microscopy, gels, blots, or other visual data, are not illustrations, but evidence. Allowing AI-generated or AI-altered data into the scientific record would undermine the credibility of the entire enterprise. The policy does allow AI-generated illustrations for explanatory purposes, provided they are disclosed and validated by the authors. This is a pragmatic recognition that AI can be a powerful tool for visual communication. FASEB also emphasizes that any AI-assisted data analysis should be clearly explained and must adhere to COPE guidelines on data fabrication. This is a crucial reminder that AI-driven analytics, while increasingly common, must be transparent and reproducible. The peer review process is one of the most vulnerable points in the publication pipeline. Unpublished manuscripts contain confidential data, proprietary methods, and early-stage findings that must be protected. FASEB's policy therefore prohibits reviewers, editors, and editorial board members from uploading manuscript content into external Gen AI tools. Doing so would violate confidentiality and potentially expose copyrighted material to third-party systems. Reviewers may use Gen AI to polish their own written feedback, but only their own text, never the manuscript itself and should disclose such use. This strikes a reasonable balance between leveraging technology for clarity and preserving the sanctity of confidential research. We likewise want to be transparent with authors, readers, and reviewers about our own use of AI, as this is an evolving space. Since the journal first launched, all manuscripts submitted undergo human review. Today, this review is supplemented by AI-based integrity screening tools operating on secure servers. We want to be clear: the AI tools we use are licensed tools, behind a secure firewall, and fully protect the author's intellectual property. We would not risk your intellectual property for the sake of expediency or convenience. We are also implementing these new tools thoughtfully, and only when we fully understand how they were developed, what they are measuring or assessing, and after a lengthy deliberation to ensure they do not breach our core values. AI tools used at FASEB are also always supplemented with human oversight, care, and deliberation; we do not make decisions about an author's submission based on an AI tool's recommendations. We will use these new tools to supplement our human decisioning by professionally trained staff, editors, and reviewers. In the same way that we ask authors to provide human oversight of their tools, we demand the same of ourselves as your publishing partners. FASEB's Gen AI policy is an affirmation of scientific values: transparency, accountability, rigor, and trust. It acknowledges the transformative potential of generative AI while insisting that human expertise remains at the center of scientific authorship and review. We hope this approach offers a model for how scientific societies can navigate the AI revolution with clarity and conviction and uphold our values. L.E.W., E.A.J. and T.H.S. all contributed to the writing, editing and final approval of the manuscript for submission. The authors acknowledge the use of Microsoft Copilot (accessed March and April 2026) to generate a bullet point summary of the major points of the FASEB Gen AI policy and create a first draft of this editorial. Microsoft Copilot (accessed April 2026) was also used as an editing tool on the text to limit redundancy and improve clarity. All AI-assisted text was reviewed and revised by the authors to ensure accuracy and clarity of meaning. The authors have nothing to report. The authors declare no conflicts of interest. No new data.