Generative Artificial Intelligence: Policy Intersections, Considerations and Recommendations

GENERATIVE ARTIFICIAL INTELLIGENCE

A guidance document on

Policy Intersections, Considerations and Recommendations

Minnesota State is an affirmative action, equal opportunity employer and educator.

Document Information

Version: 1.0

Original Publication Date: March 4, 2024

Updated: --

Questions regarding the format or substance of this document may be directed to Stephen

Kelly at stephen.kelly@minnstate.edu.

This guidance document is a product of the Minnesota State System Office and made possible

through the contributions of a multidivisional team of professionals.

Academic and Student Affairs

Stephen Kelly

Project Manager – NextGen Student

Educational Development and Technology

Kim Lynch

Associate Vice Chancellor for Educational

Development and Technology

Educational Development and Technology

Scott Wojtanowski

System Director for Educational Technology

and Development

Educational Development and Technology

Gary Hunter

System Director for Policy, Procedure, and

Intellectual Property

Student Affairs and Enrollment

Management

Catherine Ford

Program Director for Educational

Development

Educational Development and Technology

Office of the General Counsel

Daniel McCabe

Assistant General Counsel

Office of the General Counsel

Information Technology

Adam Barker

Director of M365 Services

Infrastructure Services and Operations

Mari Payton

Director of Engineering and Data

Enterprise Applications

Jim Nelson

Security Risk Analyst

Information Security

Human Resources

Paul Guillaume

HR Data Reporting Analyst

Human Resources

Office of Equity and Inclusion

Tarnjeet Kang

Director of Equity Assessment

Office of Equity and Inclusion

With additional contributions from:

Ashley Atteberry

Associate Compliance Officer

Civil Rights / Title IX Compliance

Cover Image Credit: OpenAI. (2024). ChatGPT Plus + DALL-E 3 (Jan 30 version) [Large language model]. https://chat.openai.com/chat

Contents

1 | About These Guidelines ............................................................................................................ 1

1.1 | Background......................................................................................................................... 1

1.2 | Statement of Intent ............................................................................................................ 1

1.3 | Document Updates ............................................................................................................ 1

1.4 | Definitions .......................................................................................................................... 1

2 | Existing Policies and Generative Artificial Intelligence ............................................................. 3

2.1 | Board Policy 3.26 Intellectual Property ............................................................................. 3

2.2 | Board Policy 3.27 Copyrights ............................................................................................. 3

2.3 | System Procedure 3.27.1 Copyright Clearance .................................................................. 3

2.4 | System Procedure 5.22.1 Acceptable Use of Computers and Information Technology

Resources .................................................................................................................................... 3

2.5 | Operating Instruction 5.23.2.1 Data Security Classification .............................................. 4

2.6 | Operating Instruction 5.23.3.1 Information Security Controls .......................................... 4

3 | Considerations and Recommendations .................................................................................... 5

3.1 | User Responsibility ............................................................................................................. 5

3.2 | Inappropriate Use of Generative AI ................................................................................... 5

3.2.1 | Academic Dishonesty .................................................................................................. 5

3.2.2 | Harmful Use and Misconduct...................................................................................... 6

3.3 | Syllabus Statements ........................................................................................................... 6

3.3.1 | Types of Syllabus Statements...................................................................................... 6

3.3.2 | Complementing a Syllabus Statement ........................................................................ 7

3.4 | Detecting the Use of Generative AI ................................................................................... 7

3.4.1 | AI Detection Software ................................................................................................. 8

3.4.2 | Benchmark Writing Samples ....................................................................................... 8

3.4.3 | Presence of Inaccuracies and/or Inconsistencies ....................................................... 8

3.5 | Proofreading Generative AI Outputs ................................................................................. 9

3.6 | Recognizing Bias and Inappropriate Content Generation ............................................... 10

3.7 | Ethical Considerations ...................................................................................................... 10

3.7.1 | Referencing / Citing Generative AI ............................................................................ 10

3.7.2 | Validation of Generated Data ................................................................................... 10

3.7.3 | Anonymization of Sensitive Data .............................................................................. 10

3.7.4 | Accessibility Compliance ........................................................................................... 10

3.7.5 | Required Use of Generative AI Services .................................................................... 10

3.8 | Equity Considerations ...................................................................................................... 11

4 | Procurement Process for AI Services ...................................................................................... 12

4.1 | Authorized Generative AI Services ................................................................................... 12

4.1.1 | Microsoft Copilot ...................................................................................................... 12

4.1.2 | Zoom AI Companion (aka Assistant) ......................................................................... 13

4.1.3 | Adobe Firefly ............................................................................................................. 13

5 | Learn More About Generative AI ............................................................................................ 14

5.1 | More Guidance Information ............................................................................................ 14

5.2 | Minnesota State Resources .............................................................................................. 14

5.3 | Free Learning Resources .................................................................................................. 14

5.4 | Free Classroom Materials................................................................................................. 14

Appendix A: Frequently Asked Questions ....................................................................................... i

Appendix B: Practices to Complement Syllabus Statements .......................................................... iii

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 1

1 | ABOUT THESE GUIDELINES

1.1 | Background

Generative artificial intelligence (AI)

services like ChatGPT are changing higher

education in unprecedented ways. Faculty,

staff, and students are asking important

questions regarding the acceptable use of

these services and are looking to campus

leaders for guidance. Many of these

questions are in turn directed to the

Minnesota State system office.

In September 2023 system office

representatives from Academic and Student

Affairs, Information Technology, the Office

of General Counsel, Human Resources, and

the Office of Equity and Inclusion gathered

to discuss the impact of generative AI on

higher education and the acceptable use of

generative AI services in the system. The

group reviewed existing policies and

guidelines from institutions outside the

Minnesota State system, reviewed existing

board policies and system procedures for

applicability, and from there, developed an

outline of guidance items based on

questions from faculty, staff, and

administrators in the system. This guidance

document is the result of that effort.

1.2 | Statement of Intent

This document provides guidance on the

acceptable use of generative AI services.

Campuses may use this document to inform

the development of local policies and

guidelines tailored to their campus

community. This document clarifies the

applicability of existing board policies,

system procedures, and operating

CSRC Topics | artificial intelligence. (2023, July 20). National

Institute for Standards and Technology. Retrieved Nov. 20, 2023

instructions, but is not itself a board policy,

system procedure, or operating instruction.

This document does not introduce any new

directives.

Campuses should use an equity lens to

evaluate their existing local policies and

practices for applicability and to further

clarify the parameters of acceptable

generative AI use.

The Minnesota State system office

encourages the responsible exploration and

ethical use of generative AI services. In

situations where this document does not

provide guidance to specific questions

about generative artificial intelligence,

persons are encouraged to contact system

office representatives for further guidance

through the Minnesota State Service Portal.

1.3 | Document Updates

This document will be updated periodically

in response to future events that impact the

accuracy of the information herein.

The most recent version of this document is

available on the Educational Development

and Technology webpage at minnstate.edu.

1.4 | Definitions

Artificial Intelligence (AI)

A branch of computer science devoted to

developing data processing systems that

perform functions normally associated with

human intelligence, such as reasoning,

learning, and self-improvement.

Examples: auto-captioning in Zoom,

predictive text in Microsoft Word, auto-

correct in text messaging, content

recommendations on YouTube or Facebook

from https://csrc.nist.gov/topics/technologies/artificial-

intelligence

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 2

Generative Artificial Intelligence (AI)

Artificial intelligence models that can

generate high-quality text, images, and

other content based on the data they were

trained on.

Examples: chatbots, image generators,

video generators

Large Language Models

Large language models (LLMs) are deep

learning algorithms that can recognize,

summarize, translate, predict, and generate

content using very large datasets.

These

models underpin generative artificial

intelligence services.

Examples: GPT-4, PaLM 2, Claude 2

Generative Artificial Intelligence (AI)

Service

Any cloud-based or client-side artificial

intelligence software, tool, or access point

that generates text, images, and other

content from user prompts.

Examples: ChatGPT, Google Gemini, Claude

AI, Midjourney, Microsoft Copilot

Users

Any employee or student of Minnesota

State Colleges and Universities using

computers or information technology

services for academic or business activities.

Sensitive Data

Data that is classified as highly restricted or

restricted in accordance with Operating

Instruction 5.23.2.1 Data Security

Classification.

What is Generative AI?. (2023, April 20). IBM Research Blog.

Retrieved Nov. 20, 2023 from

https://research.ibm.com/blog/what-is-generative-AI

Nvidia. (n.d.) What are Large Language Models. Retrieved Nov.

20, 2023 from https://www.nvidia.com/en-us/glossary/data-

science/large-language-models/

Bias (in generative AI)

Bias is a disproportionate weigh in favor of

or against an idea or thing, usually in a way

that is closed-minded, prejudicial, or

unfair.

Bias often appears in generative AI

outputs due to the perpetuation of bias

reflected in a model’s training data or the

bias reflected in a user’s prompt.

1: “A Distinguished Business Person.” An example of bias

in generative AI image generation. Microsoft. (2024)

Copilot with DALL-E 3 [Large language model].

https://copilot.microsoft.com.

Minnesota State Terms of Equity and Inclusion. (2023, October

18). Minnesota State Office of Equity and Inclusion. Retrieved

February 5, 2024 from

https://www.minnstate.edu/system/equity/docs/Minnesota-

State-Terms-of-Equity-and-Inclusion-10.2023.pdf

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 3

2 | EXISTING POLICIES AND

GENERATIVE ARTIFICIAL

INTELLIGENCE

2.1 | Board Policy 3.26 Intellectual

Property

Board Policy 3.26 provides guidance related

to the ownership and management of

intellectual property rights within

Minnesota State. Ownership of content

created using generative AI is presently

unsettled at the federal level; however, the

U.S. Copyright Office is currently gathering

information from stakeholders and the

public related to, among other things, the

intersection of intellectual property

ownership and the outputs of generative AI

services.

The outputs of generative AI services are

presently not subject to Board Policy 3.26

since the U.S. Copyright Office has

designated all outputs from generative AI

services as public domain.

Collections of generative AI content that

include components of human authorship

may be subject to limited intellectual

property protections.

Employees and

student are encouraged to connect with the

System Director for Policy, Procedure, and

Intellectual Property with questions

regarding potential ownership of

collections.

2.2 | Board Policy 3.27 Copyrights

Board Policy 3.27 provides guidance related

to the use of copyrighted works to further

Artificial Intelligence. (2023, August 30). U.S. Copyright Office.

Retrieved November 20, 2023 from

https://www.copyright.gov/newsnet/2023/1017.html

Generated by Artificial

Intelligence. (2023, March 16). U.S.

teaching, research, and public service at

Minnesota State colleges and universities.

Employees and students should refrain

from using copyrighted works as prompts

(in whole or in part) when using generative

AI services, except in circumstances where

express permission is provided from the

application of fair use or the TEACH Act is

present.

2.3 | System Procedure 3.27.1

Employees and students who intend to use

copyrighted (or potentially copyrighted)

materials in conjunction with generative AI

services should clear the copyright of said

materials in accordance with System

Procedure 3.27.1.

2.4 | System Procedure 5.22.1

Acceptable Use of Computers and

Information Technology Resources

System Procedure 5.22.1 applies to all users

of system information technology, which

may include generative AI services. The

procedure requires users to comply with all

applicable board policies, system

procedures, laws and regulations including

the Minnesota Government Data Practices

Act (MGDPA) and the Family Educational

Rights and Privacy Act (FERPA). As the title

implies, Acceptable Use of Computers and

Information Technology Resources

addresses proper and improper use of

information technology that is used for

academic or business activities.

https://copyright.gov/ai/ai_policy_guidance.pdf

Re: Zarya of the Dawn (Registration # VAu001480196). (2023,

February 21). U.S. Copyright Office. Retrieved November 20, 2023

from https://www.copyright.gov/docs/zarya-of-the-dawn.pdf

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 4

2.5 | Operating Instruction 5.23.2.1

Data Security Classification

Minnesota State has established three (3)

data classification levels consistent with the

MGDPA. The classification levels are highly

restricted, restricted, and low. Classifying

Minnesota State data elements, including

student and academic records, provides the

foundation to identify and apply security

controls commensurate with the

classification level. Some examples of highly

restricted, restricted and low data elements

found in the Operating Instruction 5.23.2.1

Data Security List include:

• Highly Restricted – Social Security

Numbers, personal health/medical

information, banking, or credit card

information, etc.

• Restricted – Student grades, transcripts,

class schedule, employee personal

contact information, individual

demographics including age, race,

ethnicity, gender, etc.

• Low – Data that by law is available to

the public upon request.

Providing or using highly restricted or

restricted data in any third-party

application or service, including generative

AI services, requires a contractual

agreement with the third party that

ensures adherence to data security and

data sharing protocols. These agreements

are subject to review by the Minnesota

State Office of General Counsel (or

alternatively, the Attorney General’s Office)

to ensure terms and conditions adhere to

applicable system policies and state/federal

statutes.

Third party agreements are also subject to a

data security review from the Information

Technology division of the Minnesota State

system office. This includes conducting a

risk assessment of the third-party service

and reviewing contractual clauses with the

vendor to ensure Minnesota State data is

properly protected from unauthorized

exposure and cybersecurity attacks.

Questions related to data security reviews

may be directed to your local campus IT

department. Questions related to the

review of contractual agreements, terms

and conditions may be directed to the

Office of General Counsel.

2.6 | Operating Instruction 5.23.3.1

Information Security Controls

Operating Instruction 5.23.3.1 defines the

security controls that are required to

protect Minnesota State data assets.

Requirements apply to enterprise systems

hosted by Minnesota State, campus-based

systems, and systems hosted by third

parties, including third parties that provide

generative AI services. Users accessing

generative AI services should consult their

designated campus information security

representative to clarify acceptable data

use parameters. The Office of General

Counsel or the Attorney General’s Office

can further assist with ensuring proper

Terms and Conditions are included in

generative AI contracts or end user license

agreements, including those agreements

that are open-source or zero cost.

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 5

3 | CONSIDERATIONS AND

RECOMMENDATIONS

3.1 | User Responsibility

Users are responsible for their use of

generative AI services when accessing these

services using institution or system

resources. Acceptable use is in part defined

by the terms set forth in Procedure 5.22.1,

Part 4 Responsibilities of All Users.

Institutions may further define acceptable

use of generative AI services in accordance

with local policies and procedures.

Inappropriate use of generative AI services

is subject to applicable system and

institutional policies, including but not

limited to System Procedure 5.22.1, student

codes of conduct, and course syllabus

statements.

3.2 | Inappropriate Use of

Generative AI

Student Use

In accordance with Board Policy 3.6 Student

Conduct, institutions shall establish a

student code of conduct. Institutions may

choose to further define the acceptable use

of generative AI services within their

student code of conduct. The student code

of conduct may clarify how the institution

addresses cases of inappropriate use,

academic dishonesty, and their associated

resolution processes.

Faculty should also consider including a

syllabus statement defining acceptable use

of generative AI services in their course. For

further guidance on syllabus statements,

see section 3.3 Syllabus Statements.

Employee Use

Institutions should communicate

expectations and parameters for employee

use of generative AI services. These

expectations and parameters should be

informed by applicable system policies and

procedures (as identified in this document)

along with campus policies related to

employee conduct.

3.2.1 | Academic Dishonesty

Submitting the outputs from generative AI

as one’s own work in the absence of proper

citation is plagiarism. Student use of

generative AI that violates the academic

expectations set forth in an institution’s

student code of conduct or a course

syllabus may constitute academic

dishonesty.

Whether a particular use of generative AI

constitutes academic dishonesty is

contingent upon the acceptable use

parameters established in a course syllabus.

For instance, students using generative AI

to complete a multiple-choice examination

may constitute academic dishonesty in one

course, but the same use may not

constitute academic dishonesty in another

course where a faculty member permits the

use of generative AI in assessments.

Similarly, a faculty member may not permit

any use of generative AI for essay writing in

one course, but that same faculty member

may permit the use of generative AI in

another course as an ideation or

developmental tool for essay writing.

As a final example, a faculty member may

not permit the inclusion of verbatim

outputs from generative AI in assignments

for one course, but that same faculty

member may permit properly cited

verbatim outputs from generative AI in

another course.

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 6

3.2.2 | Harmful Use and Misconduct

Generative AI services allow everyone to

create large quantities of images, audio,

and videos, and these can be used to create

deep fakes or to impersonate the likeness

of others in realistic ways. This increases

the potential for inappropriate use

appearing in the form of harmful pranks,

harassment, disinformation, and

misinformation. The Federal Bureau of

Investigation has noted a marked increase

in the number of reported incidents of

explicit content and sextortion using

generative AI.

Creating and/or disseminating the above

forms of AI generated content can be

harmful regardless of intent, and

institutions should evaluate every report of

harmful use within the scope of current

policies, including the Equal Opportunity

and Nondiscrimination in Employment and

Education (1B.1) Policy and the Sexual

Violence (1B.3, Title IX Sexual Harassment)

Policy. Institutions should consider the

information reported, conduct a close

analysis of what may be AI generated

materials, and weigh the potential impact

on campus community members.

Faculty, staff, and students should report

the harmful use of generative AI (and any

associated misconduct) to the appropriate

campus official(s) in charge of addressing

misconduct.

Federal Bureau of Investigation. (2023, June 5). Malicious Actors

Manipulating Photos and Videos to Create Explicit Content and

Sextortion Schemes. Public Service Announcement.

https://www.ic3.gov/Media/Y2023/PSA230605

Guidry, K. (2023). Syllabi policies for AI Generative Tools. Google

Docs.

2:"Pope in a Puffer Jacket." An example of deep fake

imagery. Midjourney. (2023). Midjourney (version 5).

https://midjourney.com.

3.3 | Syllabus Statements

The degree to which students are permitted

to use generative AI services to complete

course work may be specified on a course

syllabus.

3.3.1 | Types of Syllabus Statements

This section includes statements that

instructors are welcome to use in their

syllabus. These statements may be used in

part or in their entirety. Instructors are

encouraged to specify practices in their

syllabus that are congruent with their

course learning outcomes. The following

texts are adapted from the University of

Delaware and shared using a Creative

Commons license.

Considerations for Using Artificial

Intelligence in this Course

You should note that all large language

models still tend to make up incorrect facts

and fake citations. Artificial intelligence

models tend to produce inaccurate outputs,

and image generation models can

https://docs.google.com/document/d/1RMVwzjc1o0Mi8Blw_-

JUTcXv02b2WRH86vw7mi16W3U/

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 7

occasionally come up with highly offensive

products. You will be responsible for any

inaccurate, biased, offensive, or otherwise

unethical content you submit regardless of

whether it originally comes from you or an

AI tool. This syllabus includes the course

assignments and the degree to which you

may use artificial intelligence to complete

the corresponding assignment.

Practice 1: Prohibited

All work submitted during this course must

be your own. Contributions from anyone or

anything else, including AI sources, must be

properly quoted and cited every time they

are used. Failure to do so violates the

institution's academic misconduct/integrity

policy. Any allegations of academic

misconduct will be adjudicated using the

process outlined in the institution’s student

handbook.

Practice 2: Prescribed

There are situations and contexts within this

course where students will be permitted to

use generative AI tools to explore how they

can be used to complete course work. Any

student work submitted using generative AI

tools should clearly indicate what work is

the student’s work and what part is

generated by the AI. Any allegations of

academic misconduct will be adjudicated

using the process outlined in the

institution’s student handbook.

Outside of those instances that are

permitted, students are discouraged from

using generative AI tools to generate

content (text, video, audio, images) that will

end up in any student work (assignments,

activities, responses, etc.) that is used to

assess student learning.

Any allegations of academic misconduct will

be adjudicated using the process outlined in

the institution’s student handbook.

Practice 3: Open

Within this class, you are welcome to use

artificial intelligence tools in a totally

unrestricted fashion, for any purpose. Any

student work submitted using generative AI

tools should clearly indicate what work is

the student’s work and what part is

generated by the AI.

3.3.2 | Complementing a Syllabus

Statement

A syllabus statement can be a starting point

for adopting practices that complement the

acceptable use of generative AI in a course,

and faculty across Minnesota State are

actively experimenting with new practices.

Appendix B of this document contains a

growing list of practices from faculty in the

system. These practices are shared for

academic, experimental, and inspirational

purposes. These practices may not be

effective in every learning environment and

faculty should carefully evaluate each for

potential fit.

Faculty are invited to submit their practices

to the system office for potential inclusion

in this document. Contact Stephen Kelly at

stephen.kelly@minnstate.edu for more

information.

3.4 | Detecting the Use of

Generative AI

Generative AI services can generate content

that is difficult to distinguish from human

created content. This can make detecting

the use of generative AI services difficult.

Educators are searching for a singular

solution that will make detecting the use of

generative AI in academic work easy.

Unfortunately, no singular solution exists,

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 8

and due to the stochastic nature of

generative AI, there is no guarantee one will

ever emerge.

There are steps educators can take to

better detect the use of uncited generative

AI content while continuing to provide a

supportive and welcoming learning

environment for students. These steps

involve engaging students directly in

discussions about the acceptable use of AI

and the importance of referencing sources

in their work. The triangulation of multiple

indicators of generative AI use is

recommended. Educators are encouraged

to consider and explore the following

approaches to generative AI detection

alongside any methods or approaches

already in place.

3.4.1 | AI Detection Software

Since the release of generative AI services

like ChatGPT, individual developers and

companies have rushed to create tools to

detect the outputs of generative AI. The

makers of these tools boast accuracy levels

of up to 99% when used to detect writing

created by artificial intelligence.

selection of these service providers include

TurnItIn, Copyleaks, GPTZero, and

Originality AI.

Performance testing of AI detection tools

demonstrates a propensity for false positive

outcomes across the industry.

detection tools have also exhibited biased

performance when analyzing text written

by non-native English writers, with one

Desaire, H., Chua, A.E., Isom, M. et. al. (2023). Distinguishing

academic science writing from humans or ChatGPT with over 99%

accuracy using off-the-shelf machine learning tools. Cell Reports

Physical Science 4, 6. https://doi.org/10.1016/j.xcrp.2023.101426

Elkhatat, A.M., Elsaid, K. & Almeer, S. (2023). Evaluating the

efficacy of AI content detection tools in differentiating between

human and AI-generated text. Int J Educ Integr 19, 17.

https://doi.org/10.1007/s40979-023-00140-5

noteworthy study demonstrating a false

positive rate of 61.3%.

For these reasons,

educators should carefully consider use of

these tools in the analysis of student work.

The Minnesota State system office does not

recommend the use of AI detection tools as

singular indicators of plagiarism.

3.4.2 | Benchmark Writing Samples

Obtaining a benchmark writing sample from

a student can be a helpful comparative tool

when attempting to detect the use of

generative AI in academic writing. There are

a variety of ways to attain a benchmark

writing sample.

For example, one approach is to provide a

short essay question in the first weeks of a

course focused on a topic of great personal

interest to a student. This essay should

encourage a student to pull from their

personal experience and tap into their

interests, hobbies, and ambitions. This

essay question may reference course topics,

but it can also stand alone. If the class

meets in person, students can be asked to

write the essay while in class to add an

additional layer of ensured authenticity.

3.4.3 | Presence of Inaccuracies and/or

Inconsistencies

Generative AI services such as ChatGPT can

“hallucinate” in circumstances where the

user does not provide specific operating

instructions. These hallucinations typically

appear as outputs containing inaccurate

information or nonsense. The presence of

this information can be an indicator of

Sadasivan, V.S., Kumar, A., Balasubramanian, S., et. al. (2023).

Can AI-Generated Text be Reliably Detected?. Arxiv.org. Retrieved

November 22, 2023, from https://arxiv.org/pdf/2303.11156.pdf

Liang, W., Yuksekgonul, M., Mao, Y., et. al. (2023). GPT

detectors are biased against non-native English writers. Patterns

4, 7. https://doi.org/10.1016/j.patter.2023.100779

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 9

generative AI use, but educators should

take care not to confuse generative AI

hallucinations with a student’s authentic

(but uninformed) response.

3.5 | Proofreading Generative AI

Outputs

Generative AI services can produce

inaccurate, misleading, or nonsensical

content. These are commonly referred to as

“hallucinations.” The developers of these

services are actively taking steps to mitigate

the generation of inaccurate content,

however, there is no indication that

developers will be able to eliminate the

generation of inaccurate content anytime

soon.

To mitigate the impact of inaccurate

outputs, users of generative AI services may

consider the following approaches:

1. Manually review and verify the veracity

of all outputs. Do not assume accuracy.

2. Use generative AI services that provide

citations to the information provided in

outputs (e.g., Copilot from Microsoft).

Validate these citations.

3. When using generative AI services for

reasoning tasks (such as arithmetic), try

approaches like “chain-of-thought”

prompting to encourage more accurate

outputs.

4. Apply an equity lens when crafting

inputs (prompts) to reduce the

probability of inappropriate outputs.

Apply the same lens to the analysis of

outputs.

3 An example of a hallucination. OpenAI. (February 2024). ChatGPT (version 3.5). [Large language model].

https://chat.openai.com. Inspired by “Why AI Is Incredibly Smart and Shockingly Stupid.” A presentation by computer scientist

Dr. Yejin Choi.

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 10

3.6 | Recognizing Bias and

Inappropriate Content Generation

Large Language Models (the foundation of

generative AI services) are informed by

training data. This training data comes in

the form of text, and this text is pulled from

a variety of sources including books,

journals, newspapers, and much more. The

largest amount of text comes from internet

websites like Wikipedia and Reddit, as well

as data sets like Common Crawl. Because

generative AI services from American

companies are trained on text that is

representative of western cultural values,

norms, beliefs, and customs, the outputs

produced by the services are often

reflective of western-centric viewpoints.

Furthermore, these services may produce

outputs reflective of deeper societal biases

related to concepts of gender, race,

ethnicity, and other aspects of diversity.

Providers of generative AI services are

taking steps to mitigate the potential for

harmful outputs, especially those that are

offensive or disrespectful to persons

belonging to specific groups that have

historically been minoritized and are not

reflected positively or proportionally in

training data. Despite these efforts,

generative AI services still hold the

potential to produce harmful outputs. It is

important for faculty, staff, and students to

recognize the bias inherent in training data,

and to be aware of the potential for

unpredictability in outputs. Users should be

prepared to encounter inappropriate

outputs and report them to the provider of

the generative AI service accordingly.

3.7 | Ethical Considerations

3.7.1 | Referencing / Citing Generative AI

Users should always cite their use of

generative AI services in academic and

professional work consistent with standards

set forth by the American Psychological

Association (APA), the Chicago Manual of

Style, the Modern Language Association

(MLA), and similar guiding organizations.

3.7.2 | Validation of Generated Data

Users should independently validate the

outputs of generative AI services for

accuracy and fidelity to fact, especially in

circumstances where generated material

may be presented in a slidedeck, report,

application, or any other format where

observers may rely on the information for

business, academics, or research.

3.7.3 | Anonymization of Sensitive Data

In circumstances where a contract may

provide authorized use of sensitive data,

users should consider practices that protect

the subject of the data when using

generative AI services. This could include

practices such as anonymizing data and/or

restricting data access to specific roles.

3.7.4 | Accessibility Compliance

Generative AI services like ChatGPT may not

be fully accessible for those who utilize

assistive technology. Educators should be

aware of these limitations and the obstacles

they may create, and further be prepared

to pursue reasonable accommodations for

persons using assistive technology. Users

are encouraged to connect with their

campus access center for assistance.

3.7.5 | Required Use of Generative AI

Services

There may be circumstances where users

are asked or required to use generative AI

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 11

services for academics or business. It is

important to remember that generative AI

services typically use clickthrough

agreements as service gateways.

Educators cannot compel students to enter

into a contract, such as terms of use or an

end user license agreement. If a student

refuses to sign or “clickthrough” such a

contract, alternative course materials that

allow the objecting student to participate in

the course must be provided.

Faculty are encouraged to clearly note any

required use of generative AI services in

their course descriptions and course

syllabus. Students can then make an

informed enrollment decision based on the

required use.

Supervisors should consider the perspective

of employees in circumstances where the

use of a generative AI service is required.

3.8 | Equity Considerations

Equity Lens for Policy Review

Policies developed by Minnesota State

institutions to guide the use of generative

AI services should apply an equity lens to

the policy review process. This includes the

formation of a policy review team to:

• assess the purpose of the policy.

• assess who it aims to benefit and

who is left out.

• uncover assumptions.

• ensure that equity considerations

are intentional.

Policy makers are encouraged to consult

with their Campus Diversity Officer (CDO)

and/or the Office of Equity and Inclusion at

the system office for further guidance.

Equity Lens in Other Contexts

Generative AI services can support

assessments that evaluate diversity, equity,

and inclusion trends in higher education,

but when conducting statistical analyses,

researchers should control bias and

inaccuracies through the careful validation

of outputs.

Artificial intelligence can also be used to

generate audio transcriptions. This

functionality is available in select enterprise

tools in Minnesota State, including Kaltura

Mediaspace, Zoom, and Microsoft Teams.

The training data used to create these tools

typically favors “western” accents and may

not perform well in diverse participant

groups. Users are encouraged to test and

validate all transcription outputs.

Researchers are exploring the use of

generative AI in data collection and data

analysis processes. Some common use

cases include creating survey questions or

using AI to code qualitative data.

Researchers should be mindful that these

kinds of uses can embed bias into the

design of research studies, and appropriate

steps to review and validate all training and

output data from AI (while using an equity

lens) is highly recommended.

Further guidance on applying an equity lens

when conducting assessments can be

obtained from the Office of Equity and

Inclusion at the system office.

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 12

4 | PROCUREMENT PROCESS

FOR AI SERVICES

Purchasing and use of generative AI services

must follow all Minnesota State standard

procurement processes including legal,

security and contract review. This includes

zero cost services that are often obtained

via online clickthrough agreements. For

many of these ‘free’ services, the subscriber

is required to accept all terms and

conditions by clicking on an “Agree” button

to gain access to the service. System

Procedure 5.14.5 Part 3, Purchasing states:

Purchases must be prepared on forms

approved by the system office to assure that

they include all state-required contract

language. Any modifications of forms

approved by the system office or the use of

a non-system office form requires the

review by system legal counsel.

Users are encouraged to contact their

campus IT department for procurement

guidance. The system office will provide

procurement support to campuses pursuing

generative AI services for academic and

business use.

4.1 | Authorized Generative AI

Services

4.1.1 | Microsoft Copilot

The Minnesota State Microsoft 365

agreement authorizes the use of generative

AI services that Microsoft Corporation

officially provides.

Microsoft offers a suite of generative AI

services all organized under their brand

product Copilot. The Copilot ecosystem is

built upon foundational models from

OpenAI like GPT-4 and DALL-E 3. Microsoft

offers a mix of free, individual subscription,

and enterprise licensed Copilot solutions

across their client-based and cloud-based

software systems. For employees and

students, this means some Copilot

technology may be free to access (e.g.

https://copilot.microsoft.com), other

Copilot technology may require a

subscription (e.g. Microsoft 365 Copilot,

Copilot for PowerBI, Copilot Studio, etc.),

and others still may be available through

the system’s Office 365 licensing agreement

(e.g. Copilot with commercial data

protection).

Contact your campus IT department to

learn more about which Copilot services

might be available to you through the

system office or through your institution.

Copilot (Standard)

Copilot (formerly Bing Chat Enterprise) is

available in the Microsoft Edge sidebar, the

Windows operating system, and at

https://copilot.microsoft.com/. This is a

web-based AI chat tool that is built on

OpenAI’s GPT-4 and DALLE 3 models. The

consumer version of this tool is at no cost.

The Enterprise version of this tool comes

with “commercial data protection” and is

accessible by logging in with your

school/work-based Microsoft 365

credentials. (starID@minnstate.edu or

[email protected]). This tool is

included with existing Microsoft 365

licenses.

Copilot for Microsoft 365

Copilot for Microsoft 365 is an AI-powered

productivity service that coordinates large

language models, content in Microsoft

Graph, and the Microsoft 365 apps to help

users accomplish more. The estimated cost

of Copilot for Microsoft 365 is $30 per

month for each user account.

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 13

Other Copilots

Microsoft offers other Copilots, including

but not limited to:

• Copilot for PowerBI which requires

either a PowerBI Premium Capacity

(P1 or higher) or a Fabric Capacity

(F64 or higher).

• Copilot for Github to assist with

writing code

• Copilot Studio for building AI based

chatbots

• Azure AI Studio for building your

own large language models

• Microsoft security Copilot, an AI

assisted IT security tool

• Dynamics 365 Copilot

• Microsoft Sales Copilot

For technical assistance, please consult your

campus IT support staff. They can escalate

the issue to the System Office Service Desk

if necessary. Campus CIO's can also contact

the Director of M365 Services directly for

any additional inquiries.

4.1.2 | Zoom AI Companion (aka Assistant)

The Zoom AI Companion specializes at

notetaking and creating summaries of

meetings hosted in Zoom. This service is

available to all employees in Minnesota

State and approved for use through the

system’s enterprise Zoom agreement.

4.1.3 | Adobe Firefly

With simple text prompts, Adobe Firefly

allows users to generate images, add or

remove objects, transform text, and more.

Adobe Firefly is available within Adobe

Photoshop or through a web browser and is

available to users with Adobe Cloud

licenses.

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | Page 14

5 | LEARN MORE ABOUT

GENERATIVE AI

Generative AI is a rapidly evolving

technology. The Minnesota State system

office encourages students, faculty, staff,

and administrators to learn more about

how this technology works and the ways it

will continue to impact our lives inside and

outside of education. The following

resources are shared for information

purposes only. Resources and opportunities

provided from organizations outside of

Minnesota State should not be

misconstrued as representing any views,

positions, or opinions of the system.

5.1 | More Guidance Information

• State of Minnesota (MNIT) - Public

Artificial Intelligence Services

Security Standard

• US Department of Education -

Artificial Intelligence and the Future

of Teaching and Learning

• United Nations Educational,

Scientific and Cultural Organization

(UNESCO) - Guidance for generative

AI in education and research

• National Institute of Standards and

Technology (NIST) - Artificial

Intelligence Risk Management

Framework

5.2 | Minnesota State Resources

• The Network for Educational

Development - NED Team -

Generative AI Channel

• The Network for Educational

Development - NED Events Calendar

• The Network for Educational

Development – NED Community -

SharePoint Site

5.3 | Free Learning Resources

Business Users (beginner)

• Microsoft – AI for Beginners

• Google - Introduction to Generative

AI Learning Path

• Codecademy – Variety of free AI

courses

• Khan Academy – AI for Education

• Coursera – Introduction to

Generative AI

IT Professionals (beginner)

• Hugging Face - Hugging Face NLP

Course

5.4 | Free Classroom Materials

• Microsoft – Classroom Toolkit

(created for ages 13 -15, but

applicable to all learners)

• Microsoft – Prompts for EDU

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | i

APPENDIX A: FREQUENTLY ASKED QUESTIONS

Can I use ChatGPT for work related tasks?

Like social media and other free web services, institutions should maintain a list of approved

generative AI services, and this list may include ChatGPT. If an institution has approved the use

of ChatGPT, then users can use it within the parameters set by the institution. Faculty and staff

should confirm a service’s approval status prior to using it for work-related tasks.

Pursuant to Operating Instruction 5.23.2.1 Data Security Classification, no generative AI service

should be used for work involving sensitive data unless a contract is in place. If a contract is in

place (e.g. Copilot and Microsoft 365), institutional policy will determine the parameters for use

in conjunction with applicable Board Policy and Operating Procedures (Refer to section 2 of this

document).

Can students be required to use ChatGPT or similar tools?

Faculty may assign coursework that requires the use of ChatGPT or similar free tools. However,

faculty cannot compel students to enter into a contract, such as terms of use or an end user

license agreement. If a student refuses to sign or “clickthrough” such a contract, faculty must

provide alternative course materials that allow the objecting student to participate in the

course.

Faculty are encouraged to clearly note the required use of generative AI services in their course

descriptions and course syllabus so students can make an informed enrollment decision based

on the required use. If faculty assign a free generative AI service in coursework, they might

further consider alternative ways for students to satisfy the learning objectives of the course.

This will ensure that students who object to the service provider’s terms have an alternative

way to successfully learn and complete the course.

Is submitting student work to an AI detection service without a student’s

permission a violation of student copyright?

Courts have not settled this question in federal law.

In courses where detection services may be used, faculty are encouraged to include a

statement in their syllabus and the course description that indicates student work may be

submitted to an AI or plagiarism detection service.

Is submitting faculty work to a generative AI service a violation of faculty

Submitting faculty work to a generative AI service without the permission of the author may

infringe on copyrights to that work. In instances where faculty wish to allow student use of

faculty-authored course materials in generative AI prompts, they are encouraged to include the

permission in their course syllabus.

Can my institution or division purchase a license to ChatGPT for my own

professional use or use within a unit or department?

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | ii

Yes. Refer to Section 4 of this document: Procurement Process for AI Services.

When faculty or students create text, images, or other media using generative

AI, who owns them?

The U.S. Copyright Office has issued guidance designating all outputs from generative AI as

public domain. This means that no person can own the copyright to any output generated by a

generative AI service. Persons may be eligible to own the copyright to a collection that

incorporates AI generated material.

For further information, see Copyright Registration Guidance: Works Containing Material

Generated by Artificial Intelligence.

If employees use free generative AI services in the course of work, are they

assuming the liability for any violations of the services terms of use?

The answer to this question is situational. The state is required to defend employees acting

under the auspices of state business, but accepting terms of use in a personal capacity may not

qualify as state business.

Employees are encouraged to consult their campus IT department with questions related to

approved software, including free generative AI services.

If a student is suspected of using a generative AI service to complete academic

work in a prohibited way, what can be done?

See sections 3.2 Inappropriate Use of Generative AI and 3.4 Detecting the Use of Generative AI

for further guidance.

Generative Artificial Intelligence: Policy Intersections, Considerations, and Recommendations | iii

APPENDIX B: PRACTICES TO COMPLEMENT SYLLABUS STATEMENTS

The system office invites faculty to submit their practices to the system office for potential

inclusion in this document. Contact Stephen Kelly at stephen.kelly@minnstate.edu for more

information.

Practice 1

Title of Practice: Modeling how to turn off AI assistance in writing assignments.

Description: This video is a prototype of how instructors might introduce their philosophies

about AI writing assistance or generative AI tools in their courses. For example, a video or in-

class demonstration could focus on the potential for Grammarly to negatively influence a

student's thinking process. Demonstrations like this one could be created for each assignment

or group of assignments to review an instructor's guidelines and expectations. The

demonstration could be combined with other support materials for students to give them more

confidence in their own writing and thinking ability. Link to video prototype: writing with an

authentic voice.

Attribution: Kathleen Coate, Normandale Community College