University Policy 1.8
Responsible and Ethical Use of AI in Research and Scholarship

Reference Guide

1.0 Policy Introduction

The purpose of this policy is to give Cornell researchers and scholars appropriate principles for the use of AI in research and to identify the responsibilities of researchers and scholars in ensuring research involving AI is conducted responsibly, ethically, and with scientific rigor and integrity. Because AI technology is evolving rapidly, this policy addresses general principles and responsibilities. It is an interim policy and will be reviewed and updated annually until such a time as the technology and its applications become reasonably stable.

It is not the purpose of this policy to discourage the use of AI in research. The mission of Cornell University is “to discover, preserve and disseminate knowledge, to educate the next generation of global citizens, and to promote a culture of broad inquiry throughout and beyond the Cornell community.” Learning how to use AI, understanding its risks and benefits, and researching all aspects of constructing, using, and improving AI tools are encompassed within this mission.

The Cornell community has been aware of issues related to the use of AI systems for many years and efforts have been made to provide guidance and helpful tools. The website of the Cornell AI Initiative (ai.cornell.edu) serves as a compilation of resources, including the comprehensive Task Force report Generative AI in Academic Research: Perspectives and Cultural Norms, commissioned by Cornell OVPRI and completed in December of 2023. The Task Force report is the foundation on which the present policy is constructed.

2.0 To Whom This Policy Applies

The Responsible and Ethical Use of AI in Research and Scholarship Policy applies to everyone who conducts research at Cornell University on any campus.

3.0 Policy Principles

As a global leader, the Cornell research community has a significant impact on society, and Cornell researchers must avoid misuse of AI in their work. The following principles guide the ethical use of AI in research and scholarship.

3.1 Principle of Compliance

All Cornell research must be conducted legally, responsibly, and ethically. As is the case with all other research, research involving AI must be compliant with all research policies, research contract provisions, and legal and regulatory requirements. Multiple university offices routinely provide guidance and oversight to support researcher compliance.  University policies that are especially relevant when using AI are:

• University Policy 1.2, Research Integrity
• University Policy 1.4, Care and Use of Live Vertebrate Animals in Research and Teaching
• University Policy 1.5, Inventions and Related Property Rights
• University Policy 4.15, Copyright
• University Policy 4.21, Research Data Retention
• University Policy 4.22, Export and Import Control Compliance
• University Policy 4.6, Standards of Ethical Conduct
• University Policy 4.7, Retention of University Records
• University Policy 5.10, Information Security
• University Policy 8.6, Environment, Health and Safety
• WCM HRP-230.00 Human Research Protections Program Policy
• The Cornell University IRB Charge
• Cornell Privacy Principles

3.2 Principle of responsible use of data and information

Each person using an AI tool is responsible for using it responsibly and ethically, including avoiding any violations of confidentiality, copyright protection, or other legal or regulatory requirements that may apply. See section 8, Related Resources, for a link to Cornell institution-approved tools.

Cornell researchers and scholars are required to use institution-approved tools when available, avoid uploading private, export-controlled, sensitive, or proprietary data to general AI platforms, and verify the data-sharing practices embedded in these technologies in advance. No high or moderate-risk institutional data as defined in Cornell’s Policy 5.10, Information Security, may be shared with an AI system unless the relevant tool has undergone appropriate internal review. When institution-approved tools are not available, researchers and scholars must follow the IT governance process to determine an appropriate solution.

Researchers and scholars should follow Cornell’s Privacy Principles, and in particular should (i) limit use of personal or confidential data to the minimum necessary to accomplish the purpose of processing the data and (ii) take reasonable and appropriate security measures to guard against breach or misuse.

3.3 Principle of verification

It is the responsibility of each member of a research team to verify AI output and ensure it does not introduce bias or inaccuracy into the research process. This responsibility applies in all stages of research, including conception and execution of the work, dissemination of results, and translation to commercial or non-commercial use, as well as providing appropriate supervision of live interaction with AI when it is used with human participants. It also applies to the process of proposing projects, applying for funds, and writing progress reports.

Researchers leading a team (called PIs here) have a greater responsibility. The PI must ensure that all members of the team understand their responsibility in using AI and that appropriate protocols are established. For student researchers working alone or leading a project, their faculty mentors have the role of a PI regarding ensuring the students understand their responsibility in using AI.

3.4 Principle of transparency and disclosure

Consistent with disciplinary and publishing norms, the use of AI should be disclosed/identified at the time of proposing, reporting, publishing or when engaging in other dissemination activities. Use of AI to produce substantive portions of a research output and scholarly writing must be disclosed to avoid the reader mistaking that output for the work of a person or group of people, which would be misrepresentation and misattribution of intellectual contribution. Disclosure of AI use is also needed if that use is an important part of the research methodology so that scientific rigor and reproducibility can be ensured and the reader can form a fair opinion of the conclusions and strength of the results. Researchers and scholars must be mindful that the degree of transparency and disclosure required varies from field to field and should follow the norms and professional standards specific to the relevant field and discipline.

Human participant research is a case in point. If the use of AI has an impact on the study participants, which is determined by the IRB, it must be disclosed via the informed consent process. In some research projects, AI may be used to study the human ability to distinguish human output from AI-generated output, or other purposes that require the use of AI to be hidden from the participants. Research in which the use of AI must be hidden from participants requires the same consideration by the relevant IRB of ethics, informed consent, regulatory compliance, and disclosure as any other research involving deception of the participants.

4.0 Responsibilities

4.1 Responsible Office(s)

ORIA is responsible for maintaining the Compliance Guidance for AI Use in Research webpage and for bringing questions and comments submitted through the webpage to the AI in Research Policy Committee.

The AI in Research Policy Committee is responsible for the further development of this Interim University Policy and for monitoring the development of AI.

ORI is responsible for collaborating with ORIA and the AI in Research Policy Committee to further develop guidelines and to update this policy as changes in the AI landscape require.

4.2 University Management

There are no specific responsibilities of University Management related to this policy. When issues arise that, in the opinion of the AI in Research Policy Committee, need to be escalated for guidance or decision by University Management, the Committee will escalate those issues through normal channels.

4.3 Individuals

Adhering to the principles set out above is critical to using AI responsibly and ethically in research. To assist PIs and other individuals, responsibilities that reduce a substantial risk of violations are described here.

4.3.1 Verify AI Output

The responsibility for respecting confidential information, avoiding plagiarism, and otherwise conducting research and scholarship responsibly, ethically, and with scientific rigor rests with all members of the research team, but ultimately the PI is accountable. It is the responsibility of the researcher to verify information, data, and other outputs generated by any AI tool, which includes verifying the accuracy of the information and maintaining the integrity of the data. AI output related to experimental design, hazard identification, or regulatory compliance must be reviewed with the appropriate compliance office or institutional safety office.

4.3.2 Carefully consider training data and the consequences of unintended leakage

Output from AI can contain ideas and verbatim text that were used in training but not intended to be available to the user. If the nature of the research or the planned uses of AI requires using systems for which the training data is unknown or suspected of including information that is not in the public domain, PIs should follow available university guidance to avoid violation of responsible conduct of research requirements, including applicable laws and regulations. See the resources section of this policy.

4.3.3 Document and disclose AI use in research and scholarship

Reproducibility is critical to the integrity of the public research record. PIs are responsible for ensuring that the use of AI in their research is documented in publications in a manner that gives the reader sufficient understanding to successfully replicate the work. What comprises full documentation may be specified by journals or research sponsors.

While norms around the reporting of AI use are evolving, researchers are urged to err on the side of acknowledging and reporting all use of AI tools in the research process, including use in idea development, assistance with composition, code assistance, or generation of presentations and figures.

For instance, NSF currently recommends attribution in the context of proposing research: “Proposers are encouraged to indicate in the project description the extent to which, if any, AI technology was used and how it was used to develop their proposal.”

If AI is used to conduct research or scholarship that will result in a creative work with a registered copyright (e.g., exhibit, multimedia, music, script, book, dance, sculpture or other visual art form), then extra care should be taken. While the norms around authorship are evolving, if the final work contains more than de minimis AI-generated content as defined by the norms of the field of research, the nature and use of AI should be disclosed, and the parts of the work created by the author and by the AI should be distinguished and described.

Most publishers and journals require a clear attribution statement for substantive use of AI in research. The Resources section of this policy includes several publishers’ URLs for guidance on the use of AI in publishing.

PIs are also responsible for keeping a sufficiently detailed record of AI use as part of the project data or notes, and for as long as the project data is required to be retained, so that any questions of how that use impacted the results may be answered and properly disclosed for author attribution of intellectual credit, author accountability, and required declaration of AI use.

It is becoming common for an AI to be embedded in commercial software and laboratory instruments but not fully disclosed by the manufacturer. Researchers and scholars should review available information and make careful decisions on whether the particular embedded AI is appropriate for planned work. When such tools are described in research methods, the version number of the software should be included to help improve reproducibility.

4.3.4 Follow AI terms and conditions in grant contracts and publisher guidelines for AI use and attribution in research and scholarly writing

To avoid ethical and legal difficulties, sponsors are increasingly including requirements in grant contracts restricting the use of AI. It is the responsibility of researchers to avoid violating such terms, as well as adhering to terms for proper AI attribution when the use is allowed. Similarly, researchers should follow journal and publisher guidelines for reporting AI use.

4.3.5 Avoid generating patent claims with AI

Under U.S. patent law, only humans can be inventors. PIs must take care to use AI in a manner that does not result in claims in Cornell University patent applications that cannot be attributed to a human inventor. AI tools can be used to more fully describe an invention, but all claims must be the result of human, not machine, invention. PIs must document the invention process sufficiently that human inventorship is clear for all claims.

4.3.6 Avoid losing control of confidential information

If private, export-controlled, sensitive, or proprietary information, for example, student grades protected by FERPA or identifiable human participant data, is used in the research, the PI must avoid using that information as input to AI systems that have a possibility of transmitting that information outside the PI’s control. While some AI services offer a “no training use” policy as part of enterprise agreements by default, researchers should first assume that any data or information entered into commercial AI tools could become part of its training data and thus be recoverable by third parties from future iterations of the model. Commercial systems should be avoided in these cases unless the system is in a controlled environment approved by Cornell University for such use.

4.3.7 Comply with institutional safety and oversight requirements 

AI-enabled, autonomous, agentic, embodied, or generative systems used in research and scholarship are subject to the same institutional policies, oversight requirements, and regulatory obligations applicable to other research activities and technologies. The use of AI does not exempt research activities from required institutional review or oversight.

Researchers and scholars remain responsible for ensuring that AI-enabled systems are used in a manner consistent with applicable ethical, legal, regulatory, and institutional requirements. Responsibility for research conduct, participant protections, safety, and regulatory compliance may not be delegated to an AI system.

Research involving AI systems that materially influence, direct, or interact with activities involving biological agents, hazardous materials, radiation, physical hazards, human participants, or animal subjects may require review by the appropriate institutional oversight body or compliance office.

Where AI-enabled systems are used in ways that may materially affect research conduct, participant or subject welfare, laboratory operations, or safety-related decision-making, researchers are responsible for ensuring appropriate human oversight and accountability.

5.0 How to Report a Concern

Individuals are encouraged to report compliance and ethics concerns to their direct supervisor or HR representative where possible.

Where this is not feasible or where additional subject matter expertise is needed, individuals may use institutional mechanisms listed on the Concern Reporting at Cornell site.

The Cornell Ethics and Compliance Hotline is the primary mechanism to confidentially or anonymously report ethics, integrity, or compliance concerns to the University. 

6.0 Record Retention

Records associated with this policy shall be maintained by the individual or Unit engaging in the activity. Records shall be retained or disposed of in accordance with University Policy 4.7, Retention of University Records and University Policy 4.21, Research Data Retention.

7.0 Compliance

The Office of Research Integrity and Assurance for Ithaca and the Office of Research Integrity for WCM, University Compliance, University Audit, and others may audit or investigate to assess compliance with this policy. Non-compliance with university policies is addressed in accordance with applicable policies and procedures and is subject to progressive disciplinary action up to and including termination.

In addition to complying with federal, state, and local laws that apply to Cornell employees working in New York State, Cornell will also comply with applicable state and local laws in the non-NYS jurisdictions where its employees work. Therefore, if those applicable state and local laws in the non-NYS jurisdictions provide greater or additional benefits or protections not available under university policies, the university will comply with the applicable state and local laws.

8.0 Related Resources

8.1 Cornell University Resources

Cornell webpages provide further guidance and support contact information. The primary resource for AI use at Cornell is the Cornell AI Initiative.

This website contains information about AI use at Cornell, including links to AI systems that have been assessed for suitability for various purposes and others that are being evaluated.  See AI + Research for a listing of Cornell institution-approved tools.

• The primary resources for research ethics and regulatory compliance for the Ithaca, Cornell Tech, and Geneva campuses are Research Services Compliance, Ethics, and Safety webpages: Research Services
• Research ethics and regulatory compliance guidance for Weil Cornell Medicine: Research Integrity

The Compliance Guidance for AI Use in Research webpage has links to related compliance policies, AI tools approved for use at Cornell, a survey for requesting help with ethics and compliance issues related to AI use, and a survey for making comments and suggestions related to this interim policy. The AI in Research Policy Committee will review all questions, comments, and suggestions at regular meetings and may post further guidance on the website as deemed appropriate.

8.2 University and WCM Policies

• University Policy 1.2, Research Integrity
• University Policy 1.4, Care and Use of Live Vertebrate Animals in Research and Teaching
• University Policy 1.5, Inventions and Related Property Rights
• University Policy 4.15, Copyright
• University Policy 4.21, Research Data Retention
• University Policy 4.22, Export and Import Control Compliance
• University Policy 4.6, Standards of Ethical Conduct
• University Policy 4.7, Retention of University Records
• University Policy 5.10, Information Security
• University Policy 8.6, Environment, Health and Safety
• WCM Policy: WCM HRP-230.00 Human Research Protections Program Policy

8.3 Additional University Resources

• The Cornell University IRB Charge
• Cornell Privacy Principles

8.4 External Resources

• Limitations on AI use set by the Science Journals: Science Journals Editorial Policies: Artificial Intelligence
• The American Veterinary Medical Association (AVMA) maintains continuously updated guidance on AI Used in Manuscript Preparation:  AI Used in Manuscript Preparation

9.0 Definitions

10.0 Responsible Office and Policy Administration

11.0 Responsible Executive

12.0 Revision History