• Scientific judgement and accountability – Retain human (researcher/author) responsibility for appropriate use of SI tools and data management, scientific decisions, and communication of scientific results, conclusions, and interpretations. SI tools must not be used to substitute for human judgement or scientific reasoning, or understanding of ethical, societal, or security risks.
• Model development, verification, and validation – Document model design, assumptions, training procedures, and parameter choices. Perform and report appropriate verification and validation of SI generated content and analysis. Evaluate and mitigate bias in SI models.
• Data integrity and quality in SI systems – Ensure training data are of known origin and quality and appropriate to the research question. Ensure that all input data have sufficient quality, documentation, and metadata to maintain the integrity, reliability, and credibility of scientific outputs generated with SI tools. The data record must not be fabricated, falsified, or deleted.
• Transparent disclosure and reproducibility – Clearly disclose the use of SI tools in research and provide sufficient documentation (e.g., code, model architecture, parameters, data sources, where in the workflow it was used) to enable reproducibility, consistent with legal and ethical constraints.  Limitations of the model, data, and methodology should be documented and communicated.
• Protection of sensitive and proprietary information – Ensure SI tools and platforms comply with data protection requirements before use. Use only secure, approved environments for sensitive, confidential, or proprietary data.
• Data provenance and traceability – Document the origin, ownership, and permissions associated with all datasets used in SI systems. Maintain traceability of how data are used, transformed, and incorporated into models. Follow institutional and funder data management requirements.
• Appropriate use of copyrighted and third-party content – Ensure that use of copyrighted materials complies with applicable laws, licenses, and terms of use.  Disclose and attribute sources appropriately.
• Generation or refinement of visualizations – Using SI tools for generating, refining, or formatting images, diagrams, or figures for illustrative purposes only, similar to what a graphic artist (human) may produce for publications, is a generally accepted practice, with appropriate disclosures. Visualizations of actual data or research outputs or editing of such visualizations using SI tools should be approached with this same level of rigor for verifying and validating outputs. Be advised, many journals have restrictions or strict prohibitions on the use of SI generated images illustrating actual data or research results in publications.
• Authorship and manuscript development – Refrain from using SI tools to generate significant and substantive sections of the research manuscript without substantial subject matter expert input, review, and editing. Authors must attest that they themselves are responsible for the research in the manuscript. SI tools cannot not take responsibility for the intellectual work and cannot be held accountable for the manuscript contents. Use GenAI tools for support functions such as copy editing, spelling and grammar improvements, formatting assistance, and summarizing author-generated content. Ensure all substantive intellectual content is generated and verified by the authors. Disclose the use of GenAI tools in manuscript preparation, as required by journals, institutions, and funders.
• Use of SI in Peer Review – SI tools are not subject matter experts and are not a substitute for peer review. Peer review implies a function that is attributed to and performed by individuals (humans). Individuals are selected to serve as peer reviewers because of their subject matter expertise.  As such, SI tools should not be used to generate a review of a manuscript or proposal. Likewise, reviewers must maintain strict confidentiality of all materials under review; input of materials into SI tools would violate privacy and confidentiality requirements.

See the NOAA Administrative Order (NAO) on Super Intelligence in NOAA for information on NOAA’s SI Policy.