Artificial Intelligence (AI) and Research Data Management Guide Ask Us

Research Data Management (RDM) refers to the processes applied through a research project lifecycle to guide the collection, documentation, storage, sharing and preservation of research data. (“Research Data Management.” Canada.ca. Accessed August 14, 2026) The Tri-Agency expands this definition by explaining that RDM encompasses the full data lifecycle from creation, processing and analyses through preservation, storage, access, sharing and reuse - including planning, backup, dissemination and long-term preservation of research data. It also recognizes that RDM practices and definitions of research data differ across disciplines.   (“Tri‑Agency Research Data Management Policy: Frequently Asked Questions.” Canada.ca. Accessed August 14, 2026.)  Despite the disciplinary differences, the Research Data Life Cycle provides a useful framework for understanding the interconnected stages and the activities involved in managing research data throughout a project. One way to represent the Research Data Life Cycle is shown below. Artificial Intelligence (AI) can now intersect with almost every stage of the data lifecycle. At the same time, the use of AI introduces new RDM considerations including data privacy and security, provenance, reproducibility, data quality, intellectual property and preservation of AI-assisted research outputs. The table below maps some AI-related considerations to the research data lifecycle. It is intended to be a starting point, rather than an exhaustive list, for researchers who are at a stage of developing a Research Data Management Plan (DMP) for a new research project. 

Considerations for Data Management Planning in AI-assisted research across the Data Life Cycle:

FAIR data and AI

FAIR stands for: 

Findable - use sufficiently detailed metadata to identify datasets, models, and AI-generated outputs

Accessible - clearly document access restrictions and conditions

Interoperable - use appropriate standards, structured formats, controlled vocabularies, and documented schemas

Reusable - provide provenance, licensing, methodology, AI-use documentation, and sufficient contextual information

FAIR data becomes even more important when AI is involved because machines and researchers need sufficient context to interpret and reuse data. AI can process volumes of data, but it cannot independently interpret data, understand how data were collected, or assess them, whether data are trustworthy and fit for purpose.

FAIR does not mean open. Data can be FAIR while being restricted due to privacy, confidentiality, intellectual property, ethical issues, or Indigenous data governance. What matters is that sufficient information (detailed metadata) about the research data, their provenance, and accessibility is available for discovery and reuse. 

For AI-assisted research, FAIR extends beyond simply making datasets available. It is about making data understandable, contextualized, documented, and machine-actionable so that both humans and computers can use and reuse them. 

There is no single privacy framework that applies to every research project. Specific requirements may vary depending on the institution, province or territory, the type of information collected, research context and whether the research involves health information or other regulated data. This list of resources provides a useful starting point for understanding privacy requirements apply to your research. 

AI use can introduce new ethical considerations in how research involving humans is conducted, how participant information is processed, or how research findings are generated, or interpreted. Research involving humans at UNB is covered by the Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans and is overseen by the UNB Research Ethics Boards (REBs). Not every use of AI in research requires REB review. However, if AI changes how participants or their data are involved in the research, researchers need to consider whether these change needs to be discussed with the REBs.

Research involving First Nations, Metis, and Inuit peoples, communities, knowledge, languages or data requires approaches that recognize Indigenous rights and governance. Indigenous data governance encompasses sovereignty, collective rights, cultural protocols, and community authority. Indigenous research data should therefore not be treated simply as sensitive data. The Tri-Agency RDM Policy explicitly recognizes the need for distinctions-based approaches and acknowledges that Indigenous communities can guide how their data are collected, managed, used and shared.  

AI also introduces additional considerations because data can be aggregated, transformed, used to train models, or combined with other datasets in ways beyond the original research goals. These implications should be carefully evaluated and, where applicable, addressed in a data management plan (DMP), including considerations related to data governance, consent, access, use, sharing and preservation. Examples of these considerations are discussed in the section Reproducibility & Preservation. 

More resources:

• research data management planning
• sensitive data management
• Indigenous research 

The Tri-Agency RDM Policy emphasizes responsible management of research data throughout the research lifecycle, including documentation, preservation, sharing, and appropriate access. For agency-funded research, researchers should consider how AI-assisted processes affect the ability to document and preserve the data, metadata, code, and other materials needed to understand and support the research project. The current Tri-Agency policy also states that research data should be managed according to ethical, legal, commercial, and disciplinary considerations. 

Consider documenting and incorporating the following considerations into your DMP: 

AI-assisted research may produce additional materials which are important for understanding or reproducing the research. Depending on the project, these may include prompts, configuration files, code, training data, derived data and/or intermediate outputs. It is up to a researcher to determine what needs to be preserved based on the research question, disciplinary practice, ethical and legal requirements, and the information needed to understand or reproduce the research. For research subject to the Tri-Agency RDM Policy's data deposit requirement, researchers should consider whether AI-assisted data, code, metadata, or documentation directly support the research conclusions and therefore need to be appropriately managed and deposited. 

Find out more about:

• research data repositories
• UNB Dataverse - data deposit procedure
• data and code publication 
• data management planning

For any questions related to RDM, please contact RDM Services rdm.services@unb.ca 

Canadian policy and guidance

• Tri-Agency Research Data Management Policy

• Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans

• TCPS: CORE-2022 (Course on Research Ethics) 

• Privacy and the protection of personal information guidance:

  • Provincial Acts: UNB complies with New Brunswick's Right to Information and Protection of Privacy Act (RTIPPA) and Personal Health Information Privacy and Access Act (PHIPAA)

  • Federal Legislation: The Personal Information Protection and Electronic Documents Act (PIPEDA) applies to specific commercial activities

• UNB Guidelines on the use of AI

Indigenous data governance

• CARE principles

• OCAP® principles

• OCAP® library 

• First Nations Information Governance Centre resources

Research data management

• DMP guidance

• Sensitive data management

• De-identification guidance

• Research data repository guidance

AI research practices

• Artificial Intelligence Literacy

• Artificial Intelligence (AI) for Undergraduate Students

National RDM and AI Standards

CAN/DGSI 128 Standard: Developed by a national expert committee via DGSI, this first Canadian standard of its kind sets minimum requirements for organizations using AI tools to manage, preserve, and distribute research data, focusing on provenance, metadata integrity, and algorithmic transparency

Tri-Agency and National Governance: (open e-book) The Tri-Agency RDM Policy mandates Data Management Plans (DMPs) via tools like the DMP Assistant, which are increasingly adapted to account for the tracking of AI training data inputs, model lineage, and sensitive data restrictions

FASTER Framework: explores how Gen AI fits into the research lifecycle, explains how FASTER principles require projects utilizing generative AI to explicitly document accountability, security, and transparency - Using Generative AI for Research (DDN2-A63) 

Standards Council Guidance: Canadian Standards Association (CSA) Group provides active governance framework aligning with ISO/IEC 42001 to steer transparent AI policies and source data evaluation metrics for model training

Digital Research Alliance of Canada: explore the national infrastructure overview   

AI Strategy for the Federal Public Service 2025-2027: review public sector criteria