# Can I Upload Lecture Slides to ChatGPT or Another AI Tool?

> Can you upload lecture slides to ChatGPT? What four named medical school AI policies say, what private study covers in copyright law, and where the file goes.

Markdown copy of https://meduniexam.com/blog/can-i-upload-lecture-slides-to-chatgpt (the HTML page), published by MedUni Exam. Published September 20, 2026. Site guide for agents: https://meduniexam.com/llms.txt.

Written by [Nicholas](https://meduniexam.com/about.md). Checked on September 20, 2026.

Yes, ChatGPT will take the deck: upload the file and it reads it. Whether you may is a separate question, decided by your school's rule and by who owns the slides. What the tool keeps afterward is a third question, decided by a setting in your account rather than by the file. Gemini, Copilot, Claude and the study tools built on top of them raise the same three. Four US medical schools have written their answer down in public, and finding your own takes about two minutes.

## The three separate questions hiding inside "can I upload this"

One sentence carries three questions, and each has its own answer in its own place.

- **Is it allowed?** That is your school's rule, written by your own faculty, and it is the one that decides what you do on Monday morning. Look in the student handbook.
- **Is it legal?** Copyright, which belongs to whoever made the slides. Studying from a work and sending a copy of it to a company are treated differently.
- **Where does my file go?** This one is about the tool rather than about you. Any service should answer it on a page you can read before you sign up.

Blend the three and you get the answers that circulate in group chats: it is fine because everybody does it, or it is illegal because a file left your laptop. Taken apart, each has somewhere to look. Work down them in that order, because a course rule can forbid what copyright would allow, and no copyright argument gets you past a rule your school sets.

## What your school's policy probably says, in the words four of them use

Four US medical schools with public student AI policies land in nearly the same place, and I read all four on September 20, 2026. Dates below are the dates printed on the documents.

- **University of Illinois College of Medicine**, approved October 2, 2024 and revised August 5, 2026, permits students to enter college curricular material, naming lecture slides and book chapters, only into [University sanctioned firewall-protected generative AI tools](https://medicine.uic.edu/education/md/md-curriculum/educational-policies/policy-on-use-of-artificial-intelligence-ai-in-student-academic-work/).
- **[Northwestern University Feinberg School of Medicine](https://www.feinberg.northwestern.edu/md-education/docs/policies/professionalism-learning-environment/use-of-generative-ai-tools-policy.pdf)**, last approved December 4, 2024, states that students "may only upload Feinberg curricular materials (including but not limited to lecture slides, learning guides, and written feedback) into systems protected by Northwestern University IT, which requires logging in with Northwestern credentials". The policy then names one of them rather than the whole list: "Protected systems include Microsoft's large language model chatbot, Copilot in Bing." Public tools need approval from Feinberg leadership or the content authors in advance.
- **[LSU Health New Orleans School of Medicine](https://www.medschool.lsuhsc.edu/administration/docs/Policy%20on%20Use%20of%20Generative%20AI%20by%20Medical%20Students.pdf)**, last reviewed October 2025, is the blunt one: "Unless otherwise stated by the faculty member, upload of any course materials is not permitted."
- **University of Arizona College of Medicine Tucson**, [policy 4-108 dated November 13, 2024](https://www.medicine.arizona.edu/sites/default/files/2024-11/Policy-on-Student-Use-of-AI-2024-11-13.pdf), adds a setting rather than a ban: "Where available, students must opt out of having the data they submit to generative AI systems be used for the system's own training."

Same shape four times: inside the university's protected tool, yes; in a public one, not without asking. Two of the four name the person who can say yes, so asking is a real route and not a formality.

British and European schools publish fewer student-facing rules of this kind, and what does get published is usually written for a whole university rather than for a medical school. The [University of Edinburgh's guidance for students](https://information-services.ed.ac.uk/computing/comms-and-collab/elm/generative-ai-guidance-for-students/using-generative), in the version marked March 2026 and read on September 20, 2026, is the pattern: it draws its hard lines around assessed work rather than around uploads, and it steers students to ELM, the university's own platform, where your data "will not be retained by third party services to train their models or for any other purpose". The Russell Group of research universities published five [principles on July 3, 2023](https://russellgroup.ac.uk/policy/policy-briefings/principles-use-generative-ai-tools-education) that commit to teaching AI literacy and adapting assessment while leaving the detail to each institution. So the sentence that binds you is in your course handbook or your virtual learning environment.

## Who owns a lecture deck, and what private study allows

Copyright in a deck sits with whoever made it or with their employer, which on most courses means the lecturer or the university. Handing you a copy to study from is not handing you the copyright. Hence the wording in those policies about permission from the content author.

Both systems leave room for personal study. UK law has section 29 of the [Copyright, Designs and Patents Act 1988](https://www.legislation.gov.uk/ukpga/1988/48/section/29), titled Research and private study, whose subsection 1C reads: "Fair dealing with a work for the purposes of private study does not infringe any copyright in the work." American law has the fair use test in [section 107 of the Copyright Act](https://www.copyright.gov/title17/92chap1.html), four factors weighed case by case: the purpose and character of the use, the nature of the work, how much of it you used, and the effect on the market for it.

Neither line settles an upload, which surprises people. Those exceptions describe what you may do with a work; sending the file to a company adds a second party with terms of its own, and that gap is what the school policies were written to fill. None of this is legal advice, and your university library has a copyright officer whose job it is.

## Notes you wrote yourself: the file nobody disputes

Write the lecture up yourself and two of the three questions vanish. You made the file, you own it, no policy governs your own words, and a summary carries none of the figures and photographs somebody else paid to license.

Shape it like this. After a lecture, put down 400 to 600 words in full sentences, one line per mechanism, threshold, first line drug or characteristic finding. Full sentences matter because a question generator and your future self both need a claim with a verb in it. "ACE inhibitors: cough" yields nothing. "ACE inhibitors cause a dry cough through bradykinin accumulation" yields an item with a stem and a right answer.

Compressing 60 slides into 500 words is the work, and it is the part that moves recall. Summaries also carry your gaps, so questions written from them land on what you half heard rather than on whatever the slide designer gave a whole page to.

## Open licensed and government material you can upload without asking

Much of what you are taught also exists in material anyone may copy.

- **Open textbooks.** OpenStax anatomy and physiology, and the books adapted from it, carry a Creative Commons Attribution 4.0 license. Copy it, upload it, adapt it, keep the credit line.
- **US federal publications.** Under section 105 of the US Copyright Act, a work prepared by a federal officer or employee as part of their duties carries no copyright, which covers a great deal of what the National Institutes of Health publishes.
- **UK government guidance.** Public sector information such as UK Health Security Agency chapters is released under the Open Government Licence v3.0, which allows copying and adapting with attribution.
- **Drug labels.** Prescribing information published through DailyMed at the National Library of Medicine, better for pharmacology than lecture bullets.

Watch two things on that list. Attribution is a condition and not a courtesy, so keep the title, the author and the license line attached to anything you pass on. Second, an open textbook is not your syllabus: it covers your topic in its own order and at its own depth, which makes it good for mechanisms and poor for guessing what your examiner cares about.

## What happens to a file after you upload it

Three questions cover this, and a tool worth using answers all three on a page rather than in a chat reply.

- **Who receives the text?** Most study tools are a front end for one or more model providers, and that list should be written down somewhere public.
- **Does anything I send get used for training?** Paid interfaces at the large providers state that submitted content is not used to train their models. Free tiers are where this varies, which is why the Arizona policy above tells students to switch training off wherever the setting exists.
- **When is it deleted without me asking?** Retention should be a number of days, not a sentence about care.

In ChatGPT the switch has a name and a place. [OpenAI's Data Controls FAQ](https://help.openai.com/en/articles/7730893-data-controls-faq), read on September 20, 2026 and showing as updated five days earlier, gives the path: click your profile icon, select Settings, go to Data Controls, turn off "Improve the model for everyone". The page adds that "Your conversations will still appear in your chat history but won't be used to train ChatGPT." Temporary Chat is the other route, and the same page says those chats are "deleted from our systems after 30 days", are not used to train the models, are reviewed only to monitor for abuse, and are not saved in your history. That switch is what the Arizona policy is pointing at, and it takes about thirty seconds to change. Gemini and Copilot each keep their own version of it in their own account settings.

In the UK and the EU a lecture deck rarely holds personal data, so UK GDPR and the GDPR mostly do not decide this one. They start to matter the moment a case, a photograph or a patient detail is in the file, which is the line every policy above draws in its own words.

Here are this site's answers, taken from its [privacy policy](https://meduniexam.com/privacy.md) and read on September 20, 2026, with the live page carrying the current wording. Passages of your document go to OpenRouter, Groq, Cerebras, SambaNova, Mistral, Google, NVIDIA and Cloudflare Workers AI for generation, and to Anthropic for premium mode. Scanned pages go to Google Cloud Vision for text recognition. Paid interfaces state they do not train on what they receive, while some free tier providers may use submitted content to improve their own models, so a free plan job carries that risk and the app shows which lane a job will run on before it starts. Guest workspaces holding documents are deleted after 30 days with no activity, empty ones after three days, and deleting a document yourself removes its questions and results at once. [Account deletion](https://meduniexam.com/delete-account.md) says what goes and what tax law makes us keep.

One thing that paragraph does not say: this site is not a university sanctioned tool, because no university has sanctioned it. The test the four policies above set applies to it exactly as it applies to any other public tool, so a faculty slide deck does not belong here unless your school or the author of the deck says otherwise.

## Getting the same practice when the deck is off limits

Ask first. An email naming the tool, what it does with the file and why you want it is one paragraph long, and two of the four policies above point straight at the person who can answer. A yes in writing settles the question for the year.

Failing that, upload the summary you were going to write anyway plus the open licensed chapter covering the same mechanism. Items written that way print the sentence they came from, so checking one against your own note takes about ten seconds. To see the format before you decide anything, the [sample exam](https://meduniexam.com/sample.md) runs without a file of yours, and [the comparison with ChatGPT](https://meduniexam.com/compare/chatgpt.md) shows what a general assistant does with the same chapter.

An opinion, earned by watching both: the summary route gives the cleaner questions, because a summary is already filtered down to claims and a claim is what an exam item is made of. A deck gives you questions about whatever each slide contains, title pages and stock photographs included, and that is sometimes exactly the point, since the department that made those slides also writes your exam. The page on [turning lecture notes into exam questions](https://meduniexam.com/for/lecture-notes-to-exam-questions.md) makes that case, and [the slide deck page](https://meduniexam.com/for/lecture-slides.md) sets out what is read inside a .pptx and what is not. So where the deck is off limits, the summary is not a downgrade you are settling for.

## Run this checklist before you upload anything

- Search your student handbook for "artificial intelligence" and "generative". Two minutes, once a year.
- Find out whether your school offers a protected tool behind its own sign-in. Several now do, which settles the question.
- Sort the file into one of three piles: your own words, open licensed, or somebody else's teaching material.
- Third pile with no permission attached? Write the summary instead and upload that.
- Read the tool's page on where text goes and how long it is kept before you upload, not after.
- Switch training off in the settings of any consumer chat tool you study with, which in ChatGPT means the "Improve the model for everyone" toggle under Data Controls.
- Keep protected health information out of every tool that is not approved for it. Feinberg, LSU and Arizona each say so in as many words. None of them carves out a case you believe you have de-identified, so do not assume one exists.

Policies get revised. Every date on this page is the one printed on the document, read on September 20, 2026, and yours is worth checking again at the start of each academic year.

## Questions people ask

### Is it against the rules to upload lecture slides to ChatGPT?

That depends on your school, and your school has probably written it down. Four US medical schools whose policies are public allow curricular material only in a university protected tool: Illinois approved October 2, 2024 and revised August 5, 2026, Northwestern last approved December 4, 2024, LSU Health New Orleans last reviewed October 2025, and Arizona dated November 13, 2024. Search your own handbook for "generative" before you assume yours is looser.

### Are university lecture slides copyrighted?

Yes, and the copyright usually belongs to the lecturer or the university rather than to you. Being given a copy to study from is a license to study from it, not ownership. UK law protects fair dealing for private study under section 29 of the Copyright, Designs and Patents Act 1988, and US law has the four factor fair use test in section 107, but neither one decides whether you may send the file to a third party company.

### Can I upload my own notes to an AI tool instead?

Yes. Words you typed are yours, so no course policy governs them and no copyright question arises. Aim for 400 to 600 words per lecture in full sentences, because a fragment like "beta blockers: bradycardia" cannot become a question while a complete claim can. The only question left is the privacy one, which applies to every file you upload anywhere.

### Does uploading a PDF to ChatGPT train the model on it?

It depends on one setting and on which interface the tool is calling. In ChatGPT the switch is called "Improve the model for everyone" and sits in Settings under Data Controls, and a Temporary Chat is not used for training at all, per OpenAI's Data Controls FAQ read on September 20, 2026. Paid interfaces at the large providers state that submitted content is not used for training, while free tiers vary by provider. On this site the privacy policy names every provider that can receive document text and says which lane carries that risk.

### What can I upload if my medical school bans putting slides into AI tools?

Your own notes and summaries, open licensed textbooks such as the OpenStax anatomy and physiology series, US federal publications, and UK government guidance released under the Open Government Licence v3.0. Any of those covers the same mechanisms your lecture did. You can also email the lecturer for permission, which two of the four policies quoted above treat as a normal request.

## Related pages

- [Articles](https://meduniexam.com/blog.md): The article index: every post, newest first.
- [How Long to Study for MRCP Part 1: 13 Weeks](https://meduniexam.com/blog/how-long-to-study-for-mrcp-part-1.md): A week by week plan for MRCP(UK) Part 1 counted backward from the published 2027 diet dates, about four months in all, with the blueprint weights, the 45 minute block that matches paper pace, and the clinical sciences material your own notes will not cover.
- [How to Turn Notes Into Practice Questions](https://meduniexam.com/blog/how-to-turn-notes-into-practice-questions.md): A method for turning lecture notes into practice questions: the six prompt clauses that change what comes back, what published studies and this site's own counted rejection rates say goes wrong, and when to stop checking a set and rewrite the prompt instead.
- [How accurate are AI-written exam questions?](https://meduniexam.com/blog/how-accurate-are-ai-exam-questions.md): What the checks throw out, counted from real jobs over the last 30 days, and how to check an AI-written question yourself.
- [How it works](https://meduniexam.com/how-it-works.md): The verification pipeline, step by step, including what gets rejected and why.