Every student on a team is individually responsible for what gets submitted, including any section a teammate wrote with AI help. "I didn't know they used it" does not hold up under how university AI policies are actually written. The safer move is agreeing on AI ground rules as a team before the work starts, not after someone notices a section reads strangely.
Team-graded coursework raises this problem in a way solo assignments never do. A capstone team, an engineering design group, a business case competition squad, all submit one deliverable and receive one shared grade. But each member's actual process, including whether and how they used AI, lives inside private drafts, chat histories, and half finished documents that other teammates rarely see in full. That gap between shared grading and individual visibility is where most of the confusion about AI in group work actually starts, and it deserves a closer look before it turns into a real problem on a graded submission.
Most students entering a team project have already read some version of their school's academic integrity policy, at least in passing. Far fewer have thought through what that policy actually means once four or five people are contributing to the same document, on different sections, at different times, using different tools without necessarily comparing notes. A policy written with a single author in mind does not automatically translate into clear rules for a group, and most syllabi never spell out the difference. That vacuum is exactly where teams run into trouble, not because anyone set out to break a rule, but because nobody agreed in advance on what the rule actually meant for their specific project.
The Dilemma: One Grade, Individual Accountability
University guidance on this point is fairly consistent, and it does not leave much room for a team-based defense. Carnegie Mellon's academic integrity guidance for AI tools places final responsibility for submitted content on the individual student, even when a teammate generated part of that content using AI. That framing matters for group work specifically, because a shared grade does not create shared cover. If one section of a report was written by a chatbot and passed off as a teammate's original analysis, every name on the cover page is exposed to that problem, not only the person who typed the prompt. The team structure does not lower the standard each student is held to; it just makes the standard harder to see coming.
Consider a four-person business capstone team splitting a market analysis report into sections. One member takes the financial projections, another the competitor landscape, a third the executive summary, and the fourth the recommendations. If the competitor landscape section was drafted almost entirely by a chatbot and lightly edited, the other three members have no direct way of knowing that unless they read the underlying prompts or drafts, which they almost never do. Yet all four names go on the final report, and all four are answering for that section if it gets questioned.
Why "I Didn't Know" Doesn't Work
The instinct to treat ignorance as a defense makes sense on the surface. If a teammate used AI without saying so, the rest of the group genuinely did not know about it at the time. But academic integrity policies are not written around what a student knew during the process; they are written around what got submitted under that student's name. There is a second layer that makes group situations harder specifically. Vanderbilt's guidance on generative AI and academic integrity notes that individual instructors, not the university as a whole, set the specific AI policy for each course. That means the rule a team is supposed to follow is not one fixed campus-wide standard; it changes from class to class, and every student has to track it themselves across every course they are taking that term. A teammate who assumes "this was fine in my other class" is not applying an exception to the rule. They are applying an entirely different rule to the wrong assignment.
Put those two points together and the shape of the problem gets clearer. Responsibility sits with each individual regardless of who actually generated the content, and the specific rule that content had to follow is set at the course level, not campus-wide. A team that never discusses AI is effectively betting that everyone happens to be applying the same unwritten standard, across different courses with different instructors, without ever comparing notes out loud. That bet fails more often than it should, and it tends to fail quietly. Nobody finds out until a section reads oddly to a grader, a detection tool flags something, or a question comes up in office hours that nobody on the team can answer clearly.
Where the Defense Actually Breaks Down
- Academic integrity policies attach to what was submitted under your name, not to who knew what during the process of putting it together.
- A teammate's private AI use is not the same thing as an approved use, even when it seems harmless or minor in scope.
- Instructors cannot verify who wrote which section without the team disclosing it directly, so silence tends to read as agreement after the fact.
- Different courses set different AI rules under the same institution, so assuming last semester's policy still applies here is its own separate mistake.
- A shared grade does not create shared legal or academic cover; each name on the submission is judged on its own terms if a question comes up later.
What Actually Counts as "Using AI" on a Team
Part of why this stays confusing is that "AI use" covers a wide range of activity, and teams rarely define where their line sits. Running a paragraph through a grammar checker sits at one end of that range. Asking a chatbot to fix a citation format sits close to it. Generating an entire section from a short prompt and submitting it with light edits sits at the other end entirely, and most course policies treat those cases very differently even though a teammate glancing at a shared document cannot easily tell which one happened. A team that never names this range out loud tends to assume everyone is operating near the harmless end of it, which is often not a safe assumption once a deadline gets close and one member is short on time.
This is also where a lot of well-meaning teams get tripped up without realizing it. A student who would never consider having AI write their analysis might still ask it to restructure a paragraph, tighten an argument, or suggest a stronger topic sentence, and reasonably think that falls under normal editing help. Under some course policies it does. Under others, especially ones written around a strict "your own words" standard, it does not. Neither interpretation is unreasonable on its own, which is exactly why the team needs to settle it together rather than each member guessing independently and hoping the guesses line up.
Five Scenarios and Who Actually Answers for Them
Not every AI-in-group-work situation looks the same, and the right response changes with it. The table below breaks down five situations teams commonly run into, who tends to be accountable under standard university guidance, and what to actually do about each one before submission rather than after.
| Scenario | Who's Responsible | What to Do About It |
|---|---|---|
| A teammate secretly used AI for their whole section without telling the rest of the group | Every team member whose name is on the submission, including those who never touched AI directly | Ask directly before submission whether AI was used anywhere in the document, and resolve it before turning the work in, not after |
| The team agreed together, in advance, to use AI within the course's stated limits | The team collectively, though each member still answers individually if questioned | Write the agreement down somewhere, even briefly, and keep it in case the question comes up again later |
| One teammate is unsure whether AI is allowed for a task and simply doesn't ask | That teammate most directly, though the whole team shares exposure once the work is submitted together | Ask the instructor directly; a short, specific email usually settles it faster than guessing does |
| A teammate used AI to translate or paraphrase ideas from a source without citing the original | The teammate who used it primarily, but the team submitted the section collectively | Treat it as a normal citation problem: catch it in review and require a source before it goes out |
| The syllabus never explicitly mentions AI policy for this particular assignment | The team still, since silence in a syllabus is not treated as permission | Ask in writing and default to the stricter reasonable interpretation of the policy until you get an answer |
A Practical Method: Agree Before You Start
The most effective fix for all of this is not clever, it is just early. Before dividing up sections, the team should read the actual course policy together, out loud if needed, rather than paraphrasing it from memory or assuming it matches a different class from a previous term. If the policy allows AI for some tasks, brainstorming, outlining, checking grammar, and prohibits it for others, writing final analysis, generating code without disclosure, the team needs to agree on exactly where that line falls for this specific assignment, in this specific course, before anyone starts writing their section. Waiting until a teammate's paragraph reads strangely means that conversation happens under suspicion instead of under agreement, and that changes how it goes for everyone involved.
Ground rules worth setting explicitly, in writing, at the first team meeting:
- Which parts of the assignment, if any, AI is allowed to help with, based on the actual course policy rather than a guess.
- Whether the team requires disclosure between teammates even in cases where the instructor does not ask for it directly.
- What happens if someone realizes partway through that they already used AI in a way the group did not agree on beforehand.
- Who talks to the instructor if the group is genuinely unsure about a specific case, and how early that conversation happens relative to the deadline.
- How the team will handle a section that reads as clearly AI-written during a peer review pass, before it becomes the whole group's problem.
If a teammate's AI use looks like it crosses the line the group agreed on, raise it with them directly and early. A quiet, direct conversation before submission is a completely different situation from the whole team getting flagged together after the fact, and it usually resolves faster than people expect once it is actually said out loud.
This conversation goes more smoothly when it is treated as routine project logistics rather than an accusation. Most kickoff meetings already cover who owns which section and when drafts are due. Adding a two minute agenda item on AI use, phrased as "what does our syllabus actually say about this, and are we all reading it the same way," normalizes the topic before anyone has a reason to be defensive about it. Teams that raise it only after noticing a problem tend to have a harder, more charged conversation than teams that simply made it part of the standard planning process from the first meeting onward.
One thing worth doing individually, separate from whatever the team agrees on as a group: keep a private running note of what you personally used AI for and when, section by section, even if nobody on the team asks for that record. If the question of who did what on a project ever comes up after submission, that note protects you specifically, not the team as a whole, since it documents your own actual process rather than a group memory that may not agree six weeks later. This is not a case for a shared team tool, since the entire point is that it is your own account of your own work, kept separately from anything a teammate could edit or see. MemX is private by architecture, which makes it a reasonable place to keep that kind of personal reference without turning it into another shared document someone else on the team has access to.
01Can I get in trouble for a teammate's AI use I didn't know about?
Under standard university guidance, yes, since responsibility for a submission usually falls on every student whose name is on it, not only the person who used AI. That is why raising concerns before submission matters more than finding out afterward.
02Does every class have the same AI policy?
No. Individual instructors typically set their own AI rules per course, so a policy that applied in one class does not automatically carry over to another course, even within the same school or department.
03What if the syllabus never mentions AI at all?
Silence is not treated as permission. The safer approach is asking the instructor directly, in writing, and defaulting to the stricter interpretation of the assignment until you get a clear answer back.
04Should teammates have to disclose AI use to each other, even if the professor doesn't require it?
It usually is not required by the course itself, but agreeing on disclosure inside the team is one of the simplest ways to avoid one person carrying risk for work they never knew about.
05What should I do if I think a teammate used AI in a way that breaks the rules?
Raise it with them directly and early, before the assignment is submitted. A private conversation beforehand is a very different situation than the whole group getting questioned after the fact.
