Module 4 · Lesson 1

Copyright Basics in the AI Age

What does it actually mean to own an idea — and what happens when AI is the one generating them?

If a machine writes a poem, who holds the copyright?

A U.S. Copyright Office examiner reviewed an application for a graphic novel called Zarya of the Dawn. The author, Kristina Kashtanova, had used Midjourney to generate all the illustrations. The Office had initially registered the work — then revoked the image copyrights once it learned the images were AI-generated. The written text and arrangement remained protected. The images alone did not.

The ruling set a clear precedent in American law: human authorship is required for copyright protection. AI-generated content, without meaningful human creative control over the specific output, falls into the public domain the moment it is created.

Why Copyright Exists

Copyright is not primarily about money — it is about incentive. Governments grant creators a temporary monopoly on their expression so that creating new work is worth the effort. The U.S. Constitution frames it explicitly: "To promote the Progress of Science and useful Arts." The key word is expression. Ideas themselves are never copyrightable. Only specific, fixed expressions of ideas are protected.

This distinction is called the idea-expression dichotomy. Anyone can write a novel about a lonely robot learning to love. No one can copy the specific sentences Kazuo Ishiguro used in Klara and the Sun. The idea is free; the expression is protected.

What AI Changes — and What It Doesn't

AI tools generate outputs by finding statistical patterns in vast training datasets. They do not author in any legally recognized sense. Courts and copyright offices in the U.S., UK, EU, and Australia have consistently held that copyright requires a human author. That principle has not changed. What has changed is how easy it is to generate enormous volumes of content that looks like authored work.

This creates three real risks for creators who use AI:

Risk 1

No protection on AI output

If you generate an image or text purely through prompting with no substantial human creative input, you may not own it — and anyone can copy it freely.

Risk 2

Training data liability

AI models were trained on copyrighted works. Outputs that closely reproduce training data could create infringement liability for the user or the AI company.

Risk 3

Style appropriation

Prompting AI to "write like [living author]" or "draw like [working artist]" may not be illegal in itself, but it is ethically contested and commercially risky.

Safe Ground

Human selection and arrangement

Choosing which AI outputs to use, how to edit them, and how to combine them with original work can establish protectable copyright — for the human choices made.

The Three Elements Copyright Protects

Copyright protection automatically attaches when an original work is fixed in a tangible medium — you do not need to register or add a © symbol, though registration strengthens legal claims. The work must meet three tests:

Originality — the work must originate from a human author (not be copied from someone else), and must contain at least a minimal spark of creativity. A phone directory sorted alphabetically failed this test in Feist Publications v. Rural Telephone Service (1991).
Fixation — the work must exist in a stable, perceptible form. An improvised speech is not protected until recorded or transcribed. An AI chat response that is never saved is not fixed.
Expression, not idea — only the specific expression, not the underlying concept, style, genre, or fact, receives protection.

Key Ruling — Thaler v. Vidal (2022)

The U.S. Court of Appeals for the Federal Circuit confirmed that only humans can be inventors for patent purposes, rejecting attempts to list the AI system DABUS as inventor on two patents. While this is patent law, the same reasoning directly parallels copyright: authorship requires a human creative mind. The principle is settled; the edge cases are not.

What This Means for Your Work

When you use AI as part of a creative process — editing AI drafts, selecting from multiple AI outputs, combining AI-generated elements with original writing, drawing, or music — you are creating a human-AI collaborative work. The human-authored portions are protected. The AI-generated portions technically are not, though in practice distinguishing them in a finished work is often impossible and rarely tested in court for non-commercial student work.

The practical rule: the more you transform, the more you own. A lightly edited AI essay gives you weak claims. A heavily revised, restructured, and supplemented piece gives you strong ones. Understanding this spectrum is the foundation of everything else in this module.

Module 4 Core Principle

Copyright is about protecting specific human expression. AI changes the tools of creation but not the underlying legal and ethical logic. The more genuine human creative judgment you apply to a work, the stronger your claim to it — and the more you respect others' expression, the safer you are legally and ethically.

Quiz — Copyright Basics

Lesson 1 · Four questions

1. The U.S. Copyright Office refused to protect the AI-generated images in Zarya of the Dawn because:

Correct. The Office held that images generated entirely by AI do not meet the human authorship requirement, regardless of how artistic they look. The written text and arrangement — which Kashtanova authored — remained protected.

Not quite. The ruling was specifically about the requirement for human authorship. The images were denied protection because they were produced by AI, not by a human creator exercising direct creative control over the specific output.

2. The idea-expression dichotomy means that copyright protects:

Correct. Copyright never protects ideas, only the specific fixed expression of those ideas. This is why two authors can both write about the same theme without either infringing the other's copyright.

Incorrect. Copyright protects expression, not ideas. Writing an idea down does not protect the idea — it protects only that specific written expression. Anyone else can take the same idea and express it differently.

3. Which of the following would most strongly establish a human copyright claim over an AI-assisted work?

Correct. The Copyright Office's guidance and related rulings indicate that substantial human creative transformation of AI output — editing, restructuring, adding original content — creates protectable human-authored expression layered onto the AI foundation.

Not quite. While detailed prompting shows human intent, courts and the Copyright Office have focused on creative control over the specific output, not the input. What matters most is how much the human shaped and transformed the final result.

4. According to the principle established in Thaler v. Vidal (2022), which of the following is true?

Correct. The Federal Circuit confirmed that inventorship requires human mental conception. While the case is technically patent law, the reasoning directly parallels copyright: intellectual property rights require a human creative mind behind the work.

Not correct. The court explicitly rejected AI inventorship. While the case is patent law and the module notes it is not directly copyright, the same human-authorship principle governs both bodies of law consistently.

Lab 1 — Copyright Status Analyzer

Lesson 1 · Discuss copyright edge cases with your AI tutor

Your Task

You will present real or hypothetical AI-assisted creative scenarios to the AI tutor and analyze their copyright status together. Consider: Who created what? How much human judgment was applied? What is and is not protected?

Try: "I used ChatGPT to write a short story, then edited it heavily over three drafts. Who owns the final story?" — or describe your own scenario.

AI Assistant

Welcome to Lab 1. I'm here to help you think through copyright questions in AI-assisted creative work. Describe a scenario — real or hypothetical — and we'll analyze it together. Who holds the copyright? What's protected? What's not?

Module 4 · Lesson 2

Fair Use and Transformation

When borrowing becomes building — and how the law decides which is which.

Why can Saturday Night Live parody any celebrity freely while a student paper cannot paste a paragraph from a textbook?

In 1994, the Supreme Court ruled in Campbell v. Acuff-Rose Music that 2 Live Crew's parody of Roy Orbison's "Oh, Pretty Woman" was fair use — even though the group used the original melody and copied the opening lyric for commercial gain. The Court established that transformation is the single most important factor in fair use analysis. Parody must borrow from the original to comment on it. The borrowing was not incidental; it was the point.

This ruling reshaped how courts evaluate all fair use claims, making transformation — not just amount borrowed or commercial purpose — the central question.

The Four Fair Use Factors

Fair use is a U.S. legal doctrine that permits limited use of copyrighted material without permission under specific circumstances. It is not a bright-line rule — courts weigh four statutory factors, and all four must be considered together. No single factor is determinative.

Factor 1

Purpose and Character

Is the use transformative — does it add new meaning, expression, or message? Nonprofit educational and commentary uses are favored. Commercial uses face more scrutiny, but are not automatically excluded.

Factor 2

Nature of the Copied Work

Factual works (encyclopedias, news articles) receive less protection than creative works (novels, music). Using a small portion of a purely factual work is more easily fair use.

Factor 3

Amount and Substantiality

How much was taken, and was it the "heart" of the work? Copying 300 words from a novel is less serious than copying its single most famous sentence. Quantity and quality both matter.

Factor 4

Market Effect

Does the use harm the actual or potential market for the original? This is often considered the most economically significant factor. A transformative use that creates a new market is safer than one that substitutes for the original.

Real Cases: Fair Use Won and Lost

Fair Use WON — 2006

Bill Graham Archives v. Dorling Kindersley

A publisher reproduced seven Grateful Dead concert posters in thumbnail size within a 480-page biography. The Second Circuit found this transformative: the posters served a historical and biographical purpose very different from their original commercial purpose as concert promotions. Small size and historical context were key.

Transformative use upheld

Fair Use LOST — 1985

Harper & Row v. Nation Enterprises

The Nation magazine published 300 words from Gerald Ford's unpublished memoir before its official release, scooping Time magazine's paid excerpt. The Supreme Court found no fair use: even a tiny amount of copying can fail if it targets the "heart" of a work and destroys its market value. The 300 words were the most newsworthy portion of the book.

Fair use denied — market harm decisive

AI Training Case — 2023–2024

Ongoing Litigation: Authors Guild et al. v. OpenAI

Multiple lawsuits allege that training large language models on copyrighted books without license constitutes infringement. OpenAI argues training is transformative (learning patterns, not reproducing text). Plaintiffs argue it harms the market for their work and that outputs can reproduce substantial portions. These cases remain unresolved and will likely define AI fair use doctrine for a generation.

Unresolved — watching closely

Fair Use in Your AI-Assisted Work

When you create AI-assisted content, fair use becomes relevant in two directions: first, whether the AI's training constitutes infringement (you bear no liability for this — that is on the AI company); and second, whether you personally use copyrighted material in your prompts or as part of your output.

Stronger Fair Use Arguments

Quoting a few lines of a poem to analyze its themes in a critical essay. Using a screenshot of an advertisement in a media literacy presentation. Parodying a well-known work by mimicking its style to comment on it. Referencing factual content from news articles in a research synthesis.

Weaker Fair Use Arguments

Feeding an entire copyrighted book into an AI to get it summarized and then publishing the summary commercially. Using AI to reproduce a well-known song's lyrics in full as part of a blog post. Generating cover art that closely mimics a specific artist's signature style for commercial merchandise.

The Transformation Test in Practice

Ask yourself: does your use of someone else's work add new meaning, context, criticism, or commentary? Or does it simply reproduce the original for the same purpose it was originally created? The more clearly you can articulate what new value you are adding, the stronger your fair use argument becomes.

Style vs. Expression: A Critical Distinction

Style is not copyrightable. You can write a minimalist short story that reads exactly like Hemingway. You can compose a blues progression in the style of Robert Johnson. You can prompt AI to mimic a visual aesthetic. What you cannot do is copy the specific expression — the actual sentences, specific melody, particular image — that defines a copyrighted work.

This distinction gets complicated with AI because models can generate outputs that are stylistically similar but expressively distinct, or occasionally reproduce near-exact passages from training data. The safest approach: if your output could be mistaken for a specific copyrighted work, investigate before publishing.

Quiz — Fair Use and Transformation

Lesson 2 · Four questions

1. In Campbell v. Acuff-Rose, the Supreme Court ruled 2 Live Crew's parody to be fair use primarily because:

Correct. The Court held transformation to be the key factor. Parody requires borrowing from the original to work — the borrowing was itself the creative act. Commercial purpose alone does not defeat fair use if the use is sufficiently transformative.

Incorrect. The group was commercially motivated, and "Oh, Pretty Woman" was fully protected. The Court's ruling centered on transformation: parody must use the original to comment on it, making borrowing necessary and therefore excused.

2. Which fair use factor addresses whether your use substitutes for buying the original?

Correct. Factor 4 — market effect — asks whether the use harms the actual or potential market for the original work. If your use replaces purchases of the original, that weighs heavily against fair use.

Not quite. Market effect is Factor 4. It specifically asks whether your use displaces sales or licensing opportunities for the original work. This is often considered the most economically significant factor in fair use analysis.

3. Why did The Nation magazine lose the fair use case in Harper & Row v. Nation Enterprises despite using only 300 words?

Correct. Even a tiny amount of copying can defeat fair use if it targets the most valuable part of a work and harms its market. The Nation scooped a lucrative Time magazine deal by publishing the memoir's most newsworthy passages before release.

Incorrect. Neither attribution nor registration was the issue. The problem was that 300 words — though small in quantity — were the most commercially valuable portion of the book, and their publication directly destroyed a specific licensing deal with Time magazine.

4. When using AI to generate content, you are personally liable for infringement in the AI's training data if:

Correct. As an end user, you did not train the model and bear no liability for what went into its training data. The AI company's fair use arguments (or failures) are their legal problem. Your copyright concerns are about how you use the outputs.

Not quite. End users are not liable for how AI companies built their training datasets. That responsibility — and legal risk — rests with the organizations that collected and used the training data. Your copyright risks relate to how you use AI outputs, not how the model was trained.

Lab 2 — Fair Use Argument Builder

Lesson 2 · Practice making and stress-testing fair use arguments

Your Task

Describe a creative project that borrows from existing copyrighted material — with or without AI. Work with the AI tutor to build the strongest possible fair use argument, then stress-test it against the four factors.

Try: "I want to make a YouTube video essay that includes clips from movies I'm analyzing. Is that fair use?" — or describe your actual project.

Fair Use Tutor

AI Assistant

Welcome to Lab 2. Let's build your fair use argument. Tell me about a project where you're borrowing from copyrighted material — we'll analyze all four fair use factors and identify where your argument is strong and where it's vulnerable.

Module 4 · Lesson 3

Licensing, Creative Commons, and Open Content

The ecosystem of legal sharing — and how to navigate it when building with AI tools.

How do millions of creators share their work legally while keeping some rights — and why does this matter when you remix with AI?

When Wikipedia launched in 2001, it used the GNU Free Documentation License — a complex legal instrument designed for software manuals, not encyclopedia articles. In 2009, after years of negotiation, Wikipedia migrated to Creative Commons Attribution-ShareAlike (CC BY-SA). The move unlocked something extraordinary: Wikipedia's 60 million+ articles could now legally be reused, remixed, and republished by anyone, anywhere, as long as attribution was given and derivatives carried the same license. Today, Wikipedia content powers thousands of downstream projects, educational resources, and yes — AI training datasets. The licensing choice made in 2009 echoes through every tool trained on that data.

Why Licensing Matters

Every creative work is automatically copyrighted at creation. "All rights reserved" is the default. This means you need explicit permission before using, adapting, or distributing anyone else's work — unless fair use applies. Licenses are the mechanism by which creators grant that permission in advance, under defined conditions.

In the AI age, understanding licenses matters more than ever. AI image generators were trained on images scraped from the web. AI language models were trained on text from billions of sources. The legality of that training is contested, but the legality of your outputs — and the raw materials you build into your projects — is your responsibility.

Creative Commons Licenses

Creative Commons (CC) is a nonprofit that created a standardized set of free licenses allowing creators to specify exactly which rights they retain and which they grant. CC licenses layer on top of copyright — the work is still copyrighted, but the creator has pre-granted specific uses. There are six main CC licenses built from four elements:

BY — Attribution

Credit the Creator

All CC licenses require this. You must credit the original author in any use, adaptation, or distribution. This element appears in every CC license.

SA — ShareAlike

Keep It Open

If you adapt the work, your adapted version must carry the same license. This is the "viral" element that keeps open content open. Used in Wikipedia's CC BY-SA license.

NC — NonCommercial

No Commercial Use

You may use the work for non-commercial purposes only. "Commercial" is interpreted broadly — generating revenue, directly or indirectly, from the use typically triggers this restriction.

ND — NoDerivatives

No Changes Allowed

You can share the work but not adapt, remix, or transform it. Using AI to rewrite or substantially alter an ND-licensed text would violate this condition.

The Six CC Licenses at a Glance

From most permissive to most restrictive:

CC BY — Attribution only. Do anything you want, just credit the creator. The most permissive CC license. Ideal for maximum sharing.
CC BY-SA — Attribution + ShareAlike. Remix freely, but your remix must carry the same license. Used by Wikipedia, many open educational resources.
CC BY-NC — Attribution + NonCommercial. Use freely for non-commercial purposes; credit required.
CC BY-NC-SA — Attribution + NonCommercial + ShareAlike. Non-commercial use, derivatives must share alike.
CC BY-ND — Attribution + NoDerivatives. Share freely but do not alter. Commonly used by photographers and news organizations.
CC BY-NC-ND — The most restrictive CC license. Non-commercial, no derivatives, attribution required. "Look but don't touch."

Public Domain and CC0

Public domain works have no copyright restrictions at all. Works enter the public domain when their copyright expires (in the U.S., generally 70 years after the author's death for post-1928 works), when the creator explicitly releases them, or when they were created by the U.S. federal government. Many of the world's most valuable creative resources — Shakespeare's plays, Beethoven's symphonies, early jazz recordings, Leonardo's paintings — are in the public domain.

CC0 (Creative Commons Zero) is a legal tool that lets creators voluntarily waive all copyright and place their work in the public domain. NASA, many governments, and some open-data organizations use CC0 for scientific data and images. Project Gutenberg hosts over 70,000 public domain books available for any use.

AI and Open Licenses — A Current Tension

Stability AI's Stable Diffusion was trained partly on images scraped from Flickr, DeviantArt, and other platforms, including content under CC BY-NC licenses. Critics argue that using NC-licensed work to train a commercial AI model violates the NonCommercial restriction. The question has not been definitively resolved in court, but it illustrates why license terms matter even for training data — and why many artists are now restricting scraping through opt-out registries like Spawning.ai's "Have I Been Trained?" tool.

Practical Sourcing Strategy for AI-Assisted Projects

When building projects that incorporate third-party media, images, music, or text alongside AI-generated content, a clear sourcing workflow protects you legally and ethically:

Use Wikimedia Commons for images (most are CC or public domain, clearly labeled).
Use Unsplash or Pexels for photos (both have their own permissive licenses).
Use Free Music Archive or ccMixter for music, filtered by commercial use if needed.
Use Project Gutenberg or Standard Ebooks for literary texts.
Always verify the specific license terms — "free" does not always mean "commercial use allowed."
Keep records of source URLs, license types, and attribution requirements for every asset.

The Attribution Habit

Proper attribution is the core obligation in nearly every CC license. Format: Creator name, work title, source URL, license type. Example: "Photo by Alex Mead, 'Morning Fog Over the Bay,' Flickr, CC BY 2.0." Building this habit — even when not legally required — is a mark of professional creative integrity.

Quiz — Licensing and Open Content

Lesson 3 · Four questions

1. Wikipedia uses a CC BY-SA license. What does this require when you use Wikipedia content in your own published work?

Correct. CC BY-SA requires both attribution (BY) and ShareAlike (SA) — meaning your derivative work must carry the same CC BY-SA license, keeping the chain of open content going. This is the "viral" property of ShareAlike licenses.

Not quite. CC BY-SA has two requirements: attribution AND ShareAlike. The SA element means your adapted work must be released under the same license. This is different from a simple CC BY (attribution-only) license.

2. A photographer releases a landscape photo under CC BY-NC-ND. You want to use it as a background in a commercial product video. Is this permitted?

Correct. CC BY-NC-ND prohibits both commercial use (NC) and derivatives (ND). Using a photo in a commercial product video is a commercial use, which violates the NC restriction. You would need a different license or direct permission from the photographer.

Incorrect. The NC restriction explicitly prohibits commercial use — and generating revenue, directly or indirectly, is typically considered commercial. A product video for a commercial company is a commercial use regardless of whether the photo itself is modified.

3. What does CC0 allow a creator to do?

Correct. CC0 is not technically a license — it is a waiver. The creator gives up all copyright claims, effectively releasing the work into the public domain. Anyone can use it for any purpose without attribution or conditions.

Not quite. CC0 is a complete copyright waiver, not a license with conditions. The creator surrenders all rights, meaning anyone can use the work for any purpose — commercial, derivative, without attribution — with no conditions attached.

4. The criticism of AI image generators trained on CC BY-NC licensed images centers on which concern?

Correct. The argument is that using NC-licensed work to build a commercial AI product constitutes a commercial use that violates the NC condition — even if individual generated images don't reproduce the training images exactly. This legal question remains unresolved.

Not quite. CC licenses do not require disclosure or transparency — only attribution (in BY licenses) and, where applicable, license continuation (SA) or non-commercial use (NC). The specific concern is whether using NC-licensed works in commercial AI training violates the NC restriction.

Lab 3 — License Evaluator

Lesson 3 · Navigate real licensing decisions for your projects

Your Task

Describe an asset you want to use in a project (image, music, text, video clip) and tell the AI tutor what you know about its license — or where you found it. Work together to determine what you can legally do, what attribution you owe, and what alternatives exist if the license is too restrictive.

Try: "I found a great photo on Flickr for my school project. The license says CC BY-NC 2.0. Can I use it in a presentation I'm submitting for a class competition with a cash prize?" — or bring a real asset question.

License Advisor

AI Assistant

Welcome to Lab 3. Tell me about an asset you want to use — where you found it, what the license says (if you can find it), and what you plan to do with it. We'll work through whether your use is permitted and what you need to do to stay compliant.

Module 4 · Lesson 4

Ethical Remixing and Creative Integrity

Beyond what the law allows — what does it mean to be an honest creative in the age of AI remixing?

When AI can generate anything in anyone's style in seconds, what does originality even mean — and does it still matter?

In April 2023, a track called "Heart on My Sleeve" went viral on TikTok and Spotify. It featured eerily convincing AI-generated vocals mimicking Drake and The Weeknd — complete with their lyrical styles, vocal textures, and production aesthetics. The creator, Ghostwriter977, used AI voice cloning and music generation tools. Universal Music Group had the track removed, citing copyright in the artists' voices and performances. Spotify and Apple Music pulled it. The Recording Academy declared it ineligible for Grammy consideration.

The incident sparked a global debate. The track was technically impressive. Some argued it was satire or artistic commentary. Others — including many musicians — argued it was exploitation: using real artists' identities to generate content they never made, without consent, for clout and potential profit. The line between remixing and replacing had been crossed.

The Ethics Framework: Four Questions

Legal compliance is the floor, not the ceiling. A use can be technically legal and still be ethically indefensible — and vice versa, a use can be legally risky while being clearly ethical. When you remix, adapt, or build on others' creative work using AI, four questions help navigate the ethics:

Question 1

Consent and Context

Would the original creator recognize your use as legitimate? Does your work exist in a context (parody, criticism, education) that creators generally accept, or does it appropriate their identity without acknowledgment?

Question 2

Economic Harm

Does your work displace income the original creator would otherwise receive? Parody critique is generally harmless; AI-generated music "by" a real artist that competes directly with their catalog is not.

Question 3

Attribution and Honesty

Are you transparent about what is yours, what is borrowed, and what is AI-generated? Passing off AI work as entirely original human authorship is a form of deception, even when legal.

Question 4

New Value

Does your work add something meaningful — new insight, new perspective, new form — that justifies the borrowing? Or is it primarily a shortcut that extracts value from others without contributing any?

Real Cases: Ethical Lines in AI Creativity

Ethical Concern — 2022–2023

AI Art Generators and Living Artists' Styles

Artists including Greg Rutkowski, Sarah Andersen, and Kelly McKernan discovered their names were among the most-used prompts on Midjourney and Stable Diffusion — their distinctive styles being replicated by thousands of users generating commercial and hobbyist work. Rutkowski, a concept artist, noted that searches for his name returned more AI imitations than his actual work. While style itself is not copyrightable, the systematic replication of a living artist's visual identity at scale raised serious questions about economic harm and creative autonomy. Over 15,000 artists signed an open letter calling for opt-in consent for AI training.

Legally contested — ethically significant

Ethical Practice — 2023

The New York Times v. Microsoft/OpenAI (Filed December 2023)

The NYT's lawsuit alleged that ChatGPT could reproduce near-verbatim articles when prompted specifically. The Times offered evidence of outputs that reproduced hundreds of words from specific articles without attribution. Beyond the legal claim, this raised an attribution ethics question: even if AI cannot "cite its sources" automatically, users who knowingly reproduce substantial published journalism in their own work without attribution are engaging in plagiarism, regardless of how the text was generated.

Litigation ongoing — attribution ethics clear

Attribution in AI-Assisted Work

Attribution in AI-assisted creative work operates on two levels: attributing the AI tool you used, and attributing the human sources your work drew upon. Many academic institutions, publishers, and professional organizations now require disclosure of AI tool use. This is not primarily about credit — it is about honesty with your audience about how a work was made.

Honest AI Attribution Practices

"This essay was drafted with the assistance of Claude AI and substantially revised by the author." — "The images in this project were generated using Midjourney v6 based on original prompts." — "Research synthesis assisted by Perplexity AI; all sources independently verified."

Problematic Practices

Submitting AI-generated work as entirely your own in academic or professional contexts without disclosure. Generating content in a specific person's style or voice and presenting it as that person's actual work. Using AI to paraphrase copyrighted sources and presenting the result as original research.

Building an Original Voice with AI

The most important creative skill in the AI age is not prompting — it is judgment. Anyone can generate. The differentiating skill is knowing what to keep, what to discard, how to combine, what to transform, and when to start over. This judgment is what makes a work yours.

Every significant creative tradition involves building on what came before. Shakespeare adapted existing stories. Jazz emerged from blues, ragtime, and gospel. Hip-hop was built on sampling. The question was never "did you borrow?" — it was always "what did you make of it?" AI makes the borrowing effortless. That makes the transformation more important, not less.

Use AI drafts as raw material, not finished work — then apply genuine creative judgment.
Disclose AI use in any context where your audience would expect to know (academic, professional, commercial).
When borrowing from human creators, ensure you are adding new meaning or value, not just repackaging their work.
Attribute both the AI tools you used and the human sources you drew upon.
Ask: would I be comfortable if the original creator saw exactly how I used their work?

The Standard That Matters

Creative integrity is not about following rules — it is about being honest with yourself and your audience about what you made, how you made it, and whose shoulders you stood on. In a world where AI makes imitation trivially easy, genuine originality and honest attribution become more valuable, not less. That is the creative standard worth building toward.

Quiz — Ethical Remixing

Lesson 4 · Four questions

1. The "Heart on My Sleeve" AI Drake/Weeknd track was removed from platforms primarily because:

Correct. UMG asserted rights in the artists' vocal performances and identities. The track used AI voice cloning to replicate their distinctive sounds without consent or compensation — a use that raises both legal (performance rights, likeness rights) and ethical (identity appropriation) concerns.

Not quite. The removal was based on Universal Music Group's copyright and related rights claims over the artists' vocal performances and identities — not on technical quality or prior contracts. The creator was an unknown entity acting without any connection to the artists or their label.

2. Artist Greg Rutkowski's main concern about AI image generators was:

Correct. Rutkowski's documented concern was that his distinctive style was being replicated so extensively that AI imitations were drowning out his actual work in search results and client attention — an economic and visibility harm even if no individual work was copied exactly.

Not quite. Style is not copyrightable, so exact reproduction of works is a different legal claim. Rutkowski's concern was the scale of style replication: his visual identity was being used without consent to produce competing commercial and hobbyist work, displacing his own visibility and income opportunities.

3. Which of the following best describes "creative integrity" in AI-assisted work?

Correct. Creative integrity is fundamentally about honesty — transparency about process, attribution of sources and tools, and honest acknowledgment of what is borrowed versus what is genuinely original. It applies to all creative work, AI-assisted or not.

Not quite. Creative integrity does not prohibit AI use or restrict it to "hard" tasks. It is about honesty with your audience about your process, attribution of what you drew upon, and genuine human creative engagement with the material you produce.

4. Why does the module argue that genuine originality becomes MORE valuable in the AI age, not less?

Correct. When anyone can generate competent imitation instantly, what stands out is genuine creative perspective, authentic voice, and meaningful judgment about what to make and why. Scarcity drives value — and authentic human creativity becomes scarcer relative to the flood of AI-assisted imitation.

Not quite. The argument is not about legal protection or AI's current quality limits — it is about scarcity and value. When AI makes imitation effortless, the things AI cannot easily provide — genuine human perspective, authentic voice, meaningful creative judgment — become comparatively more valuable.

Lab 4 — Creative Integrity Workshop

Lesson 4 · Apply the ethics framework to your own AI-assisted creative work

Your Task

Bring a real or planned AI-assisted creative project to the AI tutor. Together, you will run it through the four ethics questions: Consent and Context, Economic Harm, Attribution and Honesty, and New Value. You will also draft appropriate attribution language for the work.

Try: "I used Midjourney to create images in the style of a famous illustrator for my school zine. How do I handle attribution, and is this ethically okay?" — or describe your own project.

Creative Ethics Advisor

AI Assistant

Welcome to Lab 4. Tell me about your AI-assisted creative project — what you made, what tools you used, and who or what you drew inspiration from. We'll run through the four ethics questions together and draft any attribution language you need to be transparent with your audience.

Module 4 Test — Remixing Without Ripping Off

15 questions · Score 80% or higher to pass

1. The U.S. Copyright Office ruled that AI-generated images in Zarya of the Dawn were not protected because:

Correct. Human authorship is a foundational requirement for copyright protection. AI-generated content alone does not meet this standard.

Incorrect. The ruling was about the human authorship requirement — a legal standard, not an aesthetic judgment.

2. Which part of Zarya of the Dawn retained copyright protection after the Copyright Office's ruling?

Correct. Human-authored text and human creative choices about arrangement and selection remained protected. Only the AI-generated images themselves were denied protection.

Incorrect. Only the human-authored portions retained protection. The AI-generated images were specifically excluded.

3. The idea-expression dichotomy means:

Correct. Ideas are free for anyone to use. Only the specific expression of those ideas — the actual words, notes, or images — receives copyright protection.

Incorrect. Ideas themselves are never copyrightable. Only the specific fixed expression of an idea is protected.

4. Campbell v. Acuff-Rose (1994) established that the most important fair use factor is:

Correct. The Supreme Court elevated transformation as the central inquiry in fair use analysis, finding that 2 Live Crew's parody was transformative even though it was commercial.

Incorrect. While commercial vs. non-commercial use is relevant, the Court established transformation — adding new meaning or expression — as the key factor.

5. In Harper & Row v. Nation Enterprises, The Nation magazine lost despite copying only 300 words because:

Correct. Quantity is less important than quality and market effect. The copied portion was the most valuable part of the memoir and its early publication directly destroyed a contracted licensing deal with Time magazine.

Incorrect. The issue was market harm combined with targeting the "heart" of the work — not commercial status, attribution, or an absolute rule about unpublished works.

6. Fair use Factor 4 — effect on the market — asks:

Correct. Factor 4 focuses on market harm to the original — does your use substitute for the original or destroy licensing opportunities? It considers both current and potential markets.

Incorrect. Factor 4 asks about harm to the original creator's market, not about the user's profit motive (that is part of Factor 1) or general financial circumstances.

7. "Style is not copyrightable" means that:

Correct. Style, genre, technique, and aesthetic approach are in the public domain. Only specific fixed expressions — the actual text, melody, or image — receive copyright protection.

Incorrect. Style not being copyrightable means the general approach is free to use, but it does not mean all uses of a living artist's identity or all copying of their specific works is permitted. It is a nuanced distinction.

8. All six Creative Commons licenses share which common element?

Correct. Attribution (BY) appears in all six CC licenses. It is the foundational requirement that all CC-licensed work shares — you must credit the creator in any use or distribution.

Incorrect. Only Attribution (BY) is present in all six CC licenses. ShareAlike, NonCommercial, and NoDerivatives are optional conditions that appear in some combinations but not all.

9. Wikipedia's CC BY-SA license means that if you adapt Wikipedia content for your own published work:

Correct. CC BY-SA is "viral" — the ShareAlike condition requires that your derivative work carry the same license, keeping the content in the open ecosystem. Attribution is also required.

Incorrect. CC BY-SA requires both attribution AND ShareAlike. Your adapted work must be released under the same license. A link alone, or the non-profit nature of Wikipedia, does not fulfill these requirements.

10. CC0 is best described as:

Correct. CC0 is a waiver, not a license. The creator surrenders all copyright claims and places the work in the public domain — anyone can use it for any purpose with no conditions, including no attribution requirement.

Incorrect. CC0 is a complete copyright waiver. The creator gives up all rights permanently, with no conditions. Attribution is not required, and the waiver cannot be revoked.

11. The "Heart on My Sleeve" AI-generated Drake/Weeknd track raised which primary ethical concern beyond its legal status?

Correct. Beyond legal questions, the ethical issue was identity appropriation — using real artists' recognizable voices and personas to create content they would never have made, without any consent, transparency, or benefit to them.

Incorrect. The core ethical concern was identity appropriation — creating content under real artists' sonic identities without consent. This goes beyond disclosure of tools or commercial status.

12. Artist Greg Rutkowski's documented concern about AI generators was primarily about:

Correct. Rutkowski documented that AI imitations of his style were flooding search results and attention, making him harder to find and competing with his actual work for commissions and visibility.

Incorrect. Style is not copyrightable, so exact reproduction claims are different. Rutkowski's concern was the economic and visibility harm from style replication at scale, not specific work copying or name commercialization.

13. Which of the following is the most ethically sound approach to attribution when publishing AI-assisted work?

Correct. Ethical attribution covers both the tools used (AI) and the human sources you built upon, in any context where your audience has a legitimate interest in knowing how the work was made.

Incorrect. Ethical disclosure applies wherever your audience has a reasonable expectation of transparency — which includes academic, professional, and often personal publishing contexts, not just commercial work.

14. According to the module, three requirements for copyright protection are originality, fixation, and expression (not idea). Which scenario fails the originality test?

Correct. The Supreme Court in Feist held that alphabetical arrangement of facts — while labor-intensive — contains no creative spark and therefore fails the originality test for copyright.

Incorrect. Common techniques (rhyme schemes, chord progressions) do not defeat originality as long as the specific expression reflects a minimal creative spark. The alphabetical directory in Feist specifically failed originality because it involved no creative arrangement choices.

15. The module argues that genuine originality becomes more valuable in the AI age because:

Correct. Scarcity drives value. When competent imitation becomes trivially easy for everyone, the things AI cannot readily provide — genuine perspective, authentic creative judgment, meaningful voice — become comparatively scarcer and more valuable.

Incorrect. The argument is not about legal protection, AI quality limits, or future regulations. It is about the economics of scarcity: when imitation becomes easy, authentic originality becomes rarer and more valuable by contrast.