Guidance on Drafting AI Use Policies

Introduction

AI tools have become a part of how work is being done: in industry, in research, and increasingly in the classroom. This statement establishes shared expectations for AI use across EECS courses so as to set clear expectations. While individual faculty will set course-specific rules, this document provides the common floor and vocabulary.

This statement was developed using the OSU College of Science Artificial Intelligence Guidelines (September 2025) as a foundational reference and adapted for the EECS context. These guidelines also align with Center for Teaching Learning (CTL) guidance on AI and the AI@OregonState guidelines. Where those documents conflict with this one, OSU policy governs. Finally, a college-wide AI policy for the College of Engineering is in development; this document will be updated to align with it when finalized.

Policy Version History

Definition

Generative AI (GenAI) is the category of AI this policy addresses. GenAI systems are built on foundation models, including large language models (LLMs), vision-language models (VLMs), and small language models, that generate human-like output such as text, code, images, audio, or other media. GenAI systems include standalone tools, agentic systems, multi-modal pipelines, and automated workflows that incorporate these models as components. Where this document refers to "AI," it means GenAI unless otherwise specified.

AI Use by Faculty

Faculty may use GenAI to create, augment, or refine instructional materials, assignments, or assessments. When they do, these responsibilities apply:

1. Faculty are required to review instructional materials created by GenAI to ensure the highest quality, accuracy, and appropriateness. If faculty choose to use AI in a substantial way to create, augment, or refine instructional materials, they are strongly encouraged to disclose in a syllabus statement their AI usage. Routine grammar assistance, search features, or autocomplete do not require disclosure.
2. Faculty may use OSU Recommended and Approved GenAI tools to provide an initial review of student work, draft feedback, or score materials against a rubric. Grades must reflect the faculty member's own evaluation of student work. Faculty must disclose in their syllabus statement when GenAI tools will be used in any part of the grading or feedback process. Only OSU Recommended and Approved tools may be used when assessing student work, as these carry the data privacy and institutional protections required for handling student records.
3. Faculty must explain their GenAI use policy in their course. Students may not recognize that AI-assisted features in common development tools constitute GenAI use under this policy. See the Student Use of AI section for the full list of tools and examples.

AI Use in Student Work

Students may use AI tools when course policies explicitly permit it. When no policy is stated, only routine grammar corrections are allowed; any other AI use must be arranged with the instructor of record before using it on coursework. Search features and autocomplete do not require arrangement or disclosure. Regardless of what AI use is permitted, students are expected to be able to explain and defend any work they submit.

The full taxonomy of permitted/prohibited use categories, with definitions, is in Appendix A. AI use exists on a spectrum from light assistance at one end to full content generation at the other. Course policies will specify which categories apply to each assignment.

AI embedded in tools you already use counts as GenAI use under this policy. GitHub Copilot, VS Code or Cursor assistants, and similar IDE features fall within this policy even when they feel more like autocomplete. When a course restricts or requires disclosure of GenAI use, that applies to embedded tools as well as standalone ones.

When you use AI on permitted work, disclose it at the point of submission: identify the tool, the type of interaction, and what portion of the final work it contributed to. For agentic tools, describe the task given and the extent of its autonomy. A one-line disclosure is sufficient in most cases. For guidance on citing AI tools in APA, MLA, and Chicago formats, see the OSU Library’s GenAI guidance.

If you are working with an external sponsor or industry partner, that engagement may carry its own AI use restrictions beyond what the syllabus states. Check with your instructor before using AI tools on sponsor-related deliverables.

Recording the instructor or other students is permitted only when the instructor allows it or as a documented disability accommodation. Oregon is an all-party consent state. Do not upload a recording of a lecture, discussion, or any other person to an AI tool without the instructor's permission.

AI Use in Research Work

This section applies to research work by EECS graduate students, undergraduate researchers, and faculty: theses, dissertations, manuscripts, grant proposals, technical reports, and documents submitted to examination committees. Where venue, sponsor, IRB, or funder rules are stricter, those rules govern.

The unit of evaluation in research is the intellectual contribution, and venues (conferences, journals, funders, committees) set their own AI rules, which are changing quickly. Treat venue rules as the primary constraint and this policy as the floor.

Permitted uses. The major advisor sets AI use norms for research conducted under their supervision. For thesis and dissertation work, AI use expectations should be set with the full committee and documented early, such as at the program of study meeting or sooner. Where no norms have been set, AI tools may be used for grammar checking, search, and autocomplete without prior approval; any substantive use requires prior approval.

Appendix B lists which uses are substantive and which require disclosure. The final product must be the researcher's, reviewed and revised to that standard. Blindly using or following AI output is itself an integrity concern.

Researcher accountability. The researcher is the author and accountable for the work.

• Verify. AI fabricates citations, misstates prior work, produces plausible but incorrect analyses, and can reproduce text from existing sources without attribution. Every factual claim, citation, quantitative result, and code path must be independently verified, and AI-assisted text should be checked for similarity to existing work.
• Own. The intellectual content (research question, design choices, analysis decisions, interpretation) must be the researcher's own. Treat AI as a tool; authorship belongs to the researcher.
• Defend. Researchers must be able to explain and justify any submitted text, analysis, or code to an advisor, committee, or reviewer.

Disclosure. Follow venue rules first. ACM, IEEE, NeurIPS, NIH, NSF, and most major venues now require disclosure of substantive AI use, and some prohibit specific uses. Where venue rules are silent, disclose substantive use in methods, acknowledgments, or an authorcontribution statement: tool, task type, and scope. Disclose any image-editing or imagegeneration process specifically. Appendix B lists which uses require disclosure, which are optional, and which need none. The same expectation applies to documents submitted to examination committees.

Confidentiality and Data handling. Do not upload sensitive or confidential material to nonapproved AI tools. This includes proprietary or sponsor-restricted material, individually identifiable information about human subjects, and other people's unpublished work, such as a coauthor's draft or a collaborator's code, used without their permission. Sponsor agreements and IRB protocols may restrict AI use further.

Self-plagiarism. Prior work published under a copyright transfer (IEEE, ACM, Elsevier, Springer) cannot be reused verbatim in later documents even when the words are the researcher's own. AI assistance does not change this. Rewrite, paraphrase with citation, or obtain permission.

It is recommended that students run a similarity check to other existing works, using Draft Coach, free via OSU Google Docs and Microsoft 365, or an equivalent tool, on all documents submitted to examination committees.

Students should escalate disagreements to the Associate Head of Graduate Programs (AHGP).

Syllabus and Assignment Statements for AI Use

The school of EECS expects every course syllabus to include an AI statement. This is a school expectation that reflects OSU’s strong encouragement and anticipated future requirement. CTL provides a GenAI syllabus statement template with tiered icons and language covering the full range from prohibited to fully permitted use. The syllabus statement should cover the Student AI use policy: what students are and are not permitted to do with AI across this course.

AI use permissions should align with course learning objectives so that grades reflect actual student mastery of those objectives. Learning outcome assessment should require independent demonstration of the underlying skill, and the AI policy for that assessment must reflect that expectation. Assignments that deviate from the course default, tighter or looser, should say so explicitly in the assignment prompt. Assignments that follow the course default need no additional statement. Refer to Appendix A for categories of permitted AI usage. (A sample syllabus table mapping these categories to course-specific examples and permissions is available in the “Resources” section.)

Syllabi must not specify grade sanctions for academic misconduct; sanction determinations belong to the College Hearing Officer. See the Academic Integrity resources for the full reporting process. Faculty who need support drafting syllabus statements or designing AI-resilient assignments can request a consultation with CTL.

Academic Integrity

Unauthorized AI use may constitute plagiarism, cheating, and / or falsification under OSU’s Code of Student Conduct.

Plagiarism: presenting AI-generated content as your own work without disclosure. Where AIgenerated text or code is incorporated into a submission, it must be attributed in the work itself. Submission-level disclosure does not substitute for in-text attribution.

Cheating: using AI to complete an assignment without prior authorization from your instructor.

Fabrication: using AI to create counterfeit citations, interview responses, or research results.

If faculty have a situation that constitutes a violation of the Code of Student Conduct and academic misconduct related to AI, faculty should follow the formal reporting process outlined in Academic Integrity resources.

AI detection tools are not reliable indicators of AI use, as stated on the CTL website, and cannot serve as sole evidence in a misconduct case. These tools have been shown to produce inaccurate results and can lead to biases in determination of culpability. A high AI probability score can encourage additional scrutiny but does not definitively prove a student used AI on an assignment. Faculty may not submit student work to AI detection tools that are not OSU Recommended and Approved, as doing so may violate institutional data privacy requirements.

Code similarity tools such as MOSS and JPlag occupy different evidentiary ground. They detect structural and textual similarity between submissions and have an established record in academic integrity cases. A MOSS flag is not proof of AI use specifically, but it is legitimate evidence of potential misconduct and may be submitted as part of a case. Faculty should not conflate code similarity findings with AI detection findings. Whether code is similar to another submission and whether a student used AI are separate questions. For ECE and other nonsoftware courses where these tools do not apply, faculty should rely on the behavioral evidence described below.

Faculty building a misconduct case should document behavioral evidence: writing or coding style inconsistent with prior submissions, responses that do not address the actual prompt, fabricated facts, fictitious references, word counts or styles inconsistent with assignment guidelines, or direct similarity to AI output when the prompt is entered into a model.

Resources

Appendix A: AI Use Categories for Coursework

The following categories define the spectrum of AI use referenced in the Student Use of AI section. These categories apply across work products: text, code, simulations, circuit designs, lab reports, and other outputs. For non-software courses, faculty should interpret and apply them in the context of their discipline.

• Clarification: using AI to explain difficult concepts, assignment prompts, or course material.
• Idea generation: brainstorming research topics, thesis statements, design concepts, or approaches.
• Summarization: condensing readings, lecture notes, or other materials to highlight key points.
• Translation: using AI to translate assignment prompts, source materials, or your own drafted text into another language; does not include submitting AI-translated text as original work without disclosure.
• Editing, proofreading, and formatting: checking grammar, tone, punctuation, and clarity within your own drafted text; creating citations, tables, and properly formatted documents; converting material into alternate formats or plain language; does not include rewriting full paragraphs.
• Coding assistance: debugging code you wrote, having AI explain an algorithm or concept, or generating sample code strictly for learning purposes; applies to software code and hardware description and specification languages.
• Rewriting or refactoring: substantively rewriting portions of text or restructuring code you authored.
• Analyzing data: using AI to generate conclusions or insights from a dataset, or to set up, run, or interpret outputs from simulation and modeling environments.
• AI-generated work product: using AI to produce the primary work submitted for an assignment, including text, code, graphics, audio, or other media; this is the highest-stakes category and is prohibited unless explicitly permitted by the instructor, or unless the generated output is itself the object of analysis rather than the submitted work product.

Appendix B: AI Use Disclosure in Research

These tiers indicate when AI use in research should be disclosed, absent stricter venue, funder, or IRB rules. They are adapted from Resnik and Hosseini (2026). Where this table and venue rules differ, venue rules govern.

Disclosure is mandatory, for example, when using AI to:

• Formulate questions or hypotheses, or design and conduct experiments.
• Draft parts of the paper, significantly revise, or synthesize text.
• Translate part or all of the paper.
• Collect, analyze, interpret, or visualize data (quantitative or qualitative).
• Extract data for a literature review and identify knowledge gaps.
• Generate synthetic data, images, or figures reported in the paper or used in the research.

Disclosure is optional, for example, when using AI to:

• Find references, or verify the relevance of references found by a person.
• Find and generate examples for existing content.
• Brainstorm or suggest how to organize a paper, or suggest a paper or section title.
• Validate or give feedback on your own existing ideas, text, or code.

Disclosure is unnecessary, for example, when using AI to:

• Check grammar, spelling, punctuation, or formatting within your own drafted text.
• Search for sources or information.
• Suggest words or phrases that improve the clarity of an existing sentence.
• Perform autocomplete, or operate as a component in a larger system where it is not generating content or making research decisions.
• Act as a digital assistant, for example to organize and maintain a project's digital assets and workflows.

Reviewer responsibilities. When reviewing manuscripts or grant proposals, do not upload the document or its contents into non-approved AI tools for review assistance. Doing so breaches the confidentiality of unpublished work. Both NSF and NIH prohibit GenAI use in its peer review process.

Additional AI Guidance