You need an illustration for a presentation, a diagram for a lab report, or a hero image for your design portfolio. Ten minutes with Nano Banana in the Gemini app and it is done. Then the questions start: who owns it, must you say it was made by AI, and what happens if your lecturer finds out you did not draw it? The answers are less scary than the rumours. Three sets of rules govern them: EU law, Google's terms, and your university's policy. Get any one of them wrong and the same image can be perfectly fine in one context and academic misconduct in another — so this guide works through all three, in plain language.
The short answer
- Yes, you can usually use AI-generated images in coursework — but only if your course allows it and you say so. Check the syllabus first.
- You probably do not own the copyright in a purely AI-generated image. EU copyright protects only the author's own intellectual creation, and a short prompt is generally not enough.
- Google's terms say your content remains yours and Google will not claim ownership of what you generate. That is a promise about Google, not a copyright certificate.
- The EU AI Act requires AI-generated images to be marked as artificial — but that duty falls mainly on Google, not on you. Your university's rules are what matter: presenting AI output as your own work is academic misconduct.
- For a portfolio, disclose the tool, show the prompt and your edits.
- Never submit a photorealistic image of a real person or place as if it were real. That is a deepfake, and the AI Act requires it to be disclosed.
What the EU AI Act actually requires
The EU AI Act, Regulation (EU) 2024/1689, has a transparency chapter — Article 50 — that has applied since 2 August 2026. It does not ban AI images or fine students: the obligations fall mostly on the companies that build and deploy AI systems, not on the person who types a prompt see article #40: The EU AI Act Explained for Students: What Actually Applies to You in 2026.
Article 50(2) says providers of AI systems that generate synthetic image, audio, video or text content must ensure the outputs are marked in a machine-readable format and detectable as artificially generated or manipulated. Article 50(4) puts a duty on deployers — the people who publish the content — to disclose it when an AI system generates or manipulates image, audio or video content that constitutes a deepfake: content that resembles existing persons, objects, places or events and would falsely appear to be authentic.
For a student, the difference is simple. A stylised illustration for a slide deck is not a deepfake: nobody would mistake a flat illustration for a photograph of a real event. A photorealistic image of a real person or event that never happened is different: if you publish it, you are the deployer, and the Act says you must disclose it.
Who owns an AI-generated image?
In the EU no statute says who owns an AI-generated image. The Court of Justice of the European Union has built up a principle instead: copyright protects only works that are the author's own intellectual creation, reflecting free and creative choices made by a human. The court confirmed this in Infopaq (C-5/08) and Painer (C-145/10): protection attaches to human creativity, not to the mere operation of a tool.
A European Parliamentary Research Service briefing from October 2026 describes the emerging consensus: purely AI-generated content with minimal human creativity does not qualify for copyright protection, while AI-assisted works may be protected where substantial human creativity was exercised — assessed case by case. This is a briefing, not a court ruling: no EU court has yet decided the copyright status of a purely AI-generated image.
The practical consequences matter more than the legal theory. If a purely generated image has no copyright, nobody owns it — not you, not Google, not the model — and you cannot stop someone else copying it. It changes when your own creativity is genuinely in the work: sketch the composition, direct the layout, iterate, then edit and finish it yourself, and the human contribution may be protected — but only your contribution, not everything the model produced. A single short prompt, on its own, is generally not a creative contribution.
What Google's terms actually say
Google's terms are the third set of rules, and the most generous. The Google Terms of Service state: "Your content remains yours, which means that you retain any intellectual property rights that you have in your content." Google also states it will not claim ownership over content you generate. Nothing in the consumer Gemini app terms bans commercial or academic use, and there is no attribution requirement.
"You retain any intellectual property rights that you have" is not the same as "you own the copyright." If the law says a purely generated image has no copyright, there is nothing for the terms to preserve. All generated images also carry an invisible SynthID watermark, so an examiner can check whether an image was made with Google AI.
The coursework line: check the syllabus before you generate
EU law and Google's terms will not get you failed. Your university can. Academic integrity rules are set course by course, and the variation is enormous. Some ban AI images outright; others treat them like stock photos; design courses often welcome them if the process is visible. Your course syllabus and your university's AI policy decide your case, so read them before you generate anything — and if they are silent, ask the lecturer in writing.
Universities that have published clear rules agree on the core point. The University of Edinburgh's student guidance says presenting AI outputs as your own work, and submitting assessments containing AI-generated images without acknowledgement, are both misconduct — while stressing that individual courses have their own rules. Anglia Ruskin University states that students must not submit work generated by AI as their own, and that unauthorised use of AI in assessments may be treated as an academic offence.
| Situation | Usually allowed? | What to do |
|---|---|---|
| Decorative illustration in slides or a report | Often yes, with disclosure | Caption it as AI-generated and name the tool |
| Diagram or figure that carries your argument | Only if the course allows it | Disclose it; keep your data visible |
| Portfolio piece for a design or media course | Often yes, and often welcomed | Show the prompt, the iterations and your edits |
| Exam, timed test or invigilated assessment | Almost never | Do not use them unless explicitly permitted |
| Photorealistic image of a real person or event | No, unless disclosed as AI-generated | Never present it as a real photo |
In an exam, the work is supposed to demonstrate your own knowledge and skills under controlled conditions, so an AI-generated image submitted there is a misrepresentation, not a shortcut. The same applies wherever the image itself is what is being marked. If the image is the skill being tested, generating it is not doing the work.
The portfolio rule: disclose the tool, show the prompt, show the edits
A portfolio is where AI images cause the least trouble — provided you are open about them. Employers and admissions tutors want evidence of how you think and work. An AI image with no context tells them nothing. The same image, presented with its prompt, three iterations, and a note about what you changed and why, shows you can direct a tool and judge its output.
A disclosure that works has four parts. First, name the tool and the model: "Image generated with Nano Banana (Gemini 2.5 Flash Image), then edited in Photoshop." Second, show the prompt, or at least the key phrases. Third, show the edits: the cropping, the colour work, the elements you redrew or replaced. Fourth, keep the human work visible: sketches, layout, typography, rationale. The AI image should be one element of a piece that is visibly yours, not the whole piece.
The risks nobody mentions
Three risks catch students out, and none of them is about the AI Act. The first is that the model can reproduce other people's work: it sometimes returns something very close to a copyrighted photograph, a trademarked character, or a recognisable brand logo. If you put that in a portfolio or a published piece, the infringement is yours, not Google's. Check for logos, watermarks and familiar characters, and regenerate anything that looks borrowed.
The second risk is people. Generating a photorealistic image of a real person without their consent can breach privacy and portrait-rights rules, and if it could be mistaken for real, it is a deepfake under the AI Act. Do not generate images of identifiable real people for coursework or portfolios at all, and never use one as evidence that something happened.
The third risk is accuracy: an AI image can look authoritative while being wrong — a diagram with mislabelled parts, a chart with invented numbers, a map with the wrong borders. If the image carries facts, check every label against a reliable source before you submit it. A wrong diagram in a lab report is not a design choice; it is a factual error, and your marker will treat it as one.
Frequently asked questions
Can I use a Nano Banana image in my coursework?
Usually yes, if your course allows it and you disclose it. Check the syllabus first: some courses ban AI images outright, and a ban in the assignment brief overrides everything else. Caption it with the tool you used, and keep the prompt and your edits.
Do I own the copyright in an image I generated?
Probably not, if the image is purely AI-generated from a short prompt. EU copyright protects works that are the author's own intellectual creation, and the emerging consensus — described in a European Parliament research briefing from October 2026 — is that purely AI-generated content with minimal human creativity does not qualify. Substantial selection, arrangement and editing may protect your human contribution, assessed case by case.
Do I have to label AI images under the EU AI Act?
The AI Act's marking duty in Article 50(2) falls on providers like Google — hence the SynthID watermark. As a student publishing content, you must disclose deepfakes under Article 50(4). But the rule that will actually affect you is your university's: presenting AI output as your own work without acknowledgement is academic misconduct see article #20: Is Using AI for Homework Cheating? Rules Every Student Should Know.
Can I put AI images in my design portfolio?
Yes — and in design courses it is often expected. Disclose the tool, show the prompt and the iterations, and show what you edited yourself. Keep the human work visible: sketches, layout, typography, rationale.
What if my course bans AI images completely?
Then do not use them, even for decoration. A course ban is a rule of the assessment, and breaking it is academic misconduct even if EU law and Google's terms would allow the image. Use stock photos or your own drawings instead.
The bottom line
AI images are a legitimate tool for students, not a loophole and not a trap. Nothing in the AI Act bans you from generating an image, and Google's terms let you use what you generate academically and commercially. But the law does not give you what most students assume it does: a purely generated image probably has no copyright, so you cannot stop others copying it, and the AI Act's labelling duties sit mostly with Google, not with you. What decides your case is your university: read the syllabus, ask when it is silent, and disclose every AI image. Do that, and the work is clearly yours see article #9: How to Write an Essay with AI: A Step-by-Step Guide for Students.
Related guides
- The EU AI Act Explained for Students: What Actually Applies to You in 2026 (article #40)
- How to Write an Essay with AI: A Step-by-Step Guide for Students (article #9)
- Is Using AI for Homework Cheating? Rules Every Student Should Know (article #20)
Facts here were checked against official pages in November 2026: Regulation (EU) 2024/1689, Article 50 (EUR-Lex); the European Parliamentary Research Service briefing on copyright and AI-generated outputs (October 2026, PE 796.995); Google's Terms of Service; and the published AI guidance of the University of Edinburgh and Anglia Ruskin University. The copyright position is grounded in the EPRS briefing and CJEU case law; a widely repeated claim about an EUIPO statement on AI-generated works could not be verified on EUIPO's own website and is not relied on here. University rules differ by course: confirm what your own syllabus allows.

