When AI Is Everywhere, What Must Students Disclose?

Article Icon Article
17 November 2026
Image by iStock/vittaya pinpan
When AI is embedded in everyday tools, meaningful disclosure policies differentiate between incidental AI assistance and uses that materially affect learning.
  • Educators should set boundaries for how students can deploy artificial intelligence tools based on the learning outcomes they want to assess.
  • Asking students to specify how AI generated or reshaped their work provides the instructor with a clearer picture of their learning than requiring them to submit generic “yes-or-no” declarations regarding the technology’s contributions.
  • When AI influences consequential academic decisions, institutions also have a responsibility to explain their use of the technology, including how they apply human oversight to AI outputs and how they give students opportunities to challenge the outcomes.

 
Recently, a simulation platform I use in one of my management courses added artificial intelligence features during a routine product update. Similarly, Copilot appears inside Word for eligible users, and Google Docs makes AI-driven phrasing suggestions as a student writes. In a recent conversation, my 14-year-old pointed out that PowerPoint Designer uses machine learning to recommend slide layouts. AI now sits inside the basic tools our students use to complete assignments, not just in deliberately opened chatbots.

This made me wonder: If students work inside platforms that now contain AI, then what exactly should they disclose about how they use it?

The common academic integrity question “Did you use AI?” assumes a distinct event that a student knowingly initiates. Yet accepting AI’s suggestions for generating a slide layout, correcting grammar, identifying and verifying sources, or creating a market analysis could all require students to produce the same answer: “Yes.”

Transparency about AI use matters. But a yes-or-no declaration neither indicates what AI contributed to an assignment nor shows whether it changed the learning evidence or crossed a boundary.

Presence and Involvement Are Not the Same

A project led by the University of Bath in the U.K. in partnership with Stellenbosch University in South Africa emphasizes AI’s increasing ubiquity. Supported by the U.K.-based Quality Assurance Agency for Higher Education, the project describes how generative AI is moving from everywhere to “inside everything.”

In her 2023 paper in the International Journal for Educational Integrity, Sarah Elaine Eaton calls this the postplagiarism era, when advanced technologies become normal parts of teaching, learning, and life. She argues that this normality makes human responsibility, attribution, and students’ ability to demonstrate learning more important, not less.

In a 2025 paper on research integrity in Accountability in Research, David Resnik and Mohammad Hosseini argue that mandatory disclosure should be reserved for intentional, substantial AI use rather than incidental AI support embedded in ordinary tools. In assessment, that distinction requires a further question: Did the contribution change what the work demonstrates?

Unless we more clearly define which AI contributions matter, students’ answers may reflect different understandings of the requirements rather than different levels of honesty.

Embedded assistance makes that question difficult to answer. Students may interpret the same declaration differently. Unless we more clearly define which AI contributions matter, their answers may reflect different understandings of the requirements rather than different levels of honesty.

A product list cannot remove this ambiguity. Platform capabilities vary by license and configuration and can change with every software update. Moreover, the same feature can matter differently across assessments. A phrase suggestion may be incidental when a finance assignment involves analysis and judgment, but material when a business communication assignment focuses on sentence-level choices. Materiality depends on the learning outcome being evaluated.

Decades ago, the internet created a similar boundary problem. In their December 2000 article in the Journal of Management Education, Constance Campbell and colleagues warn business educators about online term-paper mills. Two years later, Patrick Scanlon and David Neumann reported that some college students copied material from websites into assignments without citing the source. In its Framework for Information Literacy for Higher Education, the Association of College and Research Libraries emphasizes the acts of evaluating and attributing sources as essential to learning.

AI can obscure provenance in ways these authors did not anticipate. But their work offers a lesson that is still relevant today: Accountability should follow the contribution, not technological presence.

Ask What the Disclosure Changes

Just as the internet changed how we work and conduct research, AI is changing what disclosure can reasonably be expected to capture. Before requiring an AI declaration, we should ask this question: What decision will this disclosure help the instructor make?

In financial reporting, the International Accounting Standards Board defines information as material “if omitting, misstating, or obscuring it could reasonably be expected to influence decisions.” Similarly, as faculty determine their disclosure requirements, they need to consider how AI could be expected to influence the learning outcomes they want to assess. They should focus on what boundaries they have set, what AI contributed, how its contributions affected the evidence, and what students knowingly selected or relied on, rather than what was automatic or outside meaningful control.

When students understand these boundaries and have meaningful control over their use of AI, disclosure can serve two purposes:

It can help an instructor interpret the evidence of learning. If AI materially shaped the argument, analysis, judgment, or recommendation, that information can help determine what the work demonstrates about the student’s learning.

It can document compliance with a clearly communicated boundary or provenance requirement. AI may be restricted when students must demonstrate independent recall, writing, calculation, or another capability. Professional, confidentiality, research, or authorship requirements may also require a record of how the work was produced.

Neither purpose should be confused with a misconduct finding. A permitted contribution may still be material enough to disclose. A missing or inaccurate declaration may be relevant, but it does not by itself establish impermissible use.

For student work, the practical distinction is not between “AI” and “no AI.” It is between incidental presence and material involvement.

In a January 2026 paper in the Journal of Higher Education Policy and Management, Mark Andrew Bassett and colleagues argue that “the attempt to draw a strict line between human and AI creates more problems than it solves.” Detector scores cannot establish misconduct on their own. A fair inquiry should ask whether the boundary was clear, what AI contributed, whether it materially altered the learning evidence, what the student knowingly controlled, and whether credible evidence establishes a breach.

In other words, a declaration may inform that inquiry. It cannot decide it.

For student work, the practical distinction is therefore not between “AI” and “no AI.” It is between incidental presence and material involvement.

Let the Learning Outcome Set the Boundary

Every assessment is intended to provide evidence that students have achieved one or more learning outcomes. A case analysis may focus on whether students can exercise business judgment. A presentation may show whether they can develop and defend a strategy. A data assignment may indicate whether they can select an appropriate method, interpret the results, and explain its limitations.

In their 2024 article “Validity Matters More Than Cheating,” Phillip Dawson and co-authors argue that the deeper question is whether student work provides credible evidence of the capabilities a university claims its graduates possess.

For knowledge, students may recall ideas, explain the importance of those ideas, and make connections. For analysis and judgment, they may identify patterns, explain their approaches, justify decisions, and recognize limitations. For application, they should carry the learning into different situations.

The learning outcome should determine the AI boundary. Independent recall may need to be completed under supervised conditions without AI, such as a short in-class quiz or oral check, including when students have been approved for accommodations such as the use of assistive technology. Analysis or application may allow the use of AI as long as students explain their reasoning, defend their judgment, and respond to an alternative scenario. For example, PowerPoint Designer may be incidental when a presentation requires market analysis, but material when it is a metric for visual communication.

The AI Assessment Scale developed by Mike Perkins and colleagues offers a useful way to communicate how much AI use is permitted in evaluations. In a later commentary, this group cautioned that the scale’s implementation cannot replace necessary task redesigns nor can it make unenforceable “No AI” instruction workable. The scale cannot, by itself, show what AI contributed or whether that contribution materially changed the evidence of learning. But it can clarify expectations before students begin.

Use Targeted Disclosure, Not an Inventory

Permission and disclosure do different jobs. A boundary sets expectations in advance. Targeted disclosure identifies material involvement afterward. Even when an instructor permits the use of AI, the assignment prompt should ask whether and how AI generated or reshaped substantive content or completed expected work. Disclosure requirements should stem from assessment design, not stop at a generic integrity declaration.

For example, a student might state the following: “I used generative AI to develop initial market segmentation alternatives. I then evaluated those alternatives against the case data, selected the criteria used in the final analysis, and revised the recommendation.”

That information helps an instructor interpret the evidence. By contrast, requiring students to disclose where AI automated formatting or assisted with typographical correction may tell an instructor little about what a student has learned.

Disclosure requirements should stem from assessment design, not stop at a generic integrity declaration.

Consider a strategy simulation that will show how students interpret a changing market and justify their response. AI could be permitted during preparation. Students would then submit a recommendation, defend its assumptions, and respond to a modified scenario. Those actions provide evidence of learning and transfer. Targeted disclosure would identify where AI reshaped the analysis and what the student verified or revised.

We can look to recent commentaries for examples of frameworks that might help guide this process. A July 2026 report by Tawnya Means identifies the stages at which AI assisted the work and the intellectual decisions that human authors retained. In a 2026 AACSB Insights article, Bert Verhoeven describes a Person–Process–Product model that examines who learned, how the work was produced, and what was created. Targeted disclosure complements these approaches without turning the use of every background feature into a compliance exercise.

Match Institutional Transparency to Consequence

For their part, institutions are not exempt from their own disclosure policies. Schools may require students to account for their material choices within their knowledge and control. This requirement, however, does not discharge a school’s responsibility to explain to students when and how it uses AI in decisions that affect them.

Neither side needs to disclose every background feature. Course-level notice may suffice for low-stakes AI-assisted feedback reviewed by an instructor. If AI is involved in determining a student’s grade, raising an integrity concern, or influencing a consequential academic decision, the school should explain the technology’s role and limits, provide human review, and allow the student to question the result.

UNESCO’s 2021 Recommendation on the Ethics of Artificial Intelligence takes a proportionate approach. It states that “the level of transparency and explainability should always be appropriate to the context and impact.” With that understanding, people should know when a decision is AI-informed, and responsibility should remain attributable to the actors involved.

What Business Schools Can Change Now

Business schools can make four practical changes to refine and strengthen their AI disclosure policies:

  • State the learning outcome. Tell students what knowledge, reasoning, judgment, creation, or application they must demonstrate. Accept different forms of demonstration depending on the capability being assessed.
  • Set the AI boundary by outcome. Explain whether AI is required, permitted, limited, or restricted for each relevant component of a project and why.
  • Replace blanket declarations with targeted disclosure. Ask students to identify material AI involvement in substantive content or work outputs intended to demonstrate relevant capabilities, not to list every incidental AI-enabled feature.
  • Match institutional notice to consequence. Give general notice when AI is used in low-stakes tasks. When AI influences consequential academic decisions, explain specifically how it was used, make a person accountable for the decision, and give students a clear way to appeal the outcome to a human reviewer.

As AI becomes part of every institution’s ordinary educational infrastructure, merely inventorying its presence will produce unreliable information. The more consequential task is to identify when its involvement changes how we interpret learning, whether a boundary was respected, or how people were affected.

We should not require disclosure in every aspect of a project simply because AI was there. We should make AI involvement visible only when it impacts a learning outcome or educational decision in a material way.

Authors
Garima Kamboj
Faculty Member, Yorkville University
The views expressed by contributors to AACSB Insights do not represent an official position of AACSB, unless clearly stated.
Sign up for AACSB's LINK email newsletter.
Our subscribers receive leading insights, news, and knowledge in business education.