L1
·
Quiz
·
Lab
L2
·
Quiz
·
Lab
L3
·
Quiz
·
Lab
L4
·
Quiz
·
Lab
Module Test
Lesson 1 · Module 2

Copyright Was Not Built for This

How a 19th-century legal framework is struggling to handle outputs that no human hand directly created.
Can a machine be an author — and if not, who is?

The U.S. Copyright Office issued a landmark ruling in the case of Zarya of the Dawn — a graphic novel written by Kristina Kashtanova and illustrated using Midjourney. The office granted copyright to the text and the arrangement of images Kashtanova chose, but cancelled protection for every individual AI-generated image. The images, it concluded, were not the product of human authorship and therefore could not be owned.

The Human Authorship Requirement

American copyright law, codified in Title 17 of the U.S. Code, has required human authorship since at least 1884, when the Supreme Court decided Burrow-Giles Lithographic Co. v. Sarony. That case established that a photograph of Oscar Wilde could be copyrighted because the photographer made creative choices — lighting, pose, expression. The law has never extended protection to works produced by nature, animals, or machines acting autonomously.

The Copyright Office's 1973 COMPENDIUM (updated continuously since) states flatly: "The Office will not register works produced by nature, animals, or plants." The 2023 guidance added AI systems to that list when the AI's contribution is the creative content itself.

This creates an immediate puzzle for creators using tools like Midjourney, DALL-E, Stable Diffusion, or ChatGPT: the more expressive the AI's output, the less likely you can own it.

Documented Case

Thaler v. Perlmutter (D.D.C. 2023): Computer scientist Stephen Thaler applied to register a painting called "A Recent Entrance to Paradise," listing his AI system DABUS as the sole author. Federal District Court Judge Howell upheld the Copyright Office's refusal, writing that "human authorship is a bedrock requirement of copyright." Thaler's appeal is ongoing as of mid-2025.

What Copyright Actually Protects

Copyright protects original works of authorship fixed in a tangible medium. "Original" means it originated from a human author and shows at least a minimal degree of creativity — the bar is low, but it is not zero and it requires a human.

What copyright does NOT require: that you wrote every word or drew every stroke. A film director doesn't hold a camera or compose music, yet controls the copyright in the film. An architect's copyright covers the building design even if contractors did all the physical work. The creative decisions — what to include, how to arrange, which elements to emphasize — constitute authorship.

This distinction matters enormously for AI users. If you make the expressive choices and use AI as a tool (like using Photoshop's filters), you likely have a copyright claim. If you type a five-word prompt and press Enter, the Copyright Office's current position is that you probably don't.

Likely Protectable

Selecting and arranging AI-generated pieces into a collage · Writing surrounding text · Editing and transforming raw AI outputs · Choosing which of 200 generations to include and how to sequence them · Adding original artistic elements on top of AI output

Likely Not Protectable

The raw image from a short text prompt · The verbatim story ChatGPT wrote when you asked for one · A song generated entirely by Suno from a one-line description · Any output where the AI made all the creative choices autonomously

Why This Matters for You

If your AI-generated work lacks copyright protection, it enters the public domain immediately. Anyone can copy, sell, or remix it without your permission and without paying you. This has significant consequences for creators trying to monetize AI-assisted work — illustrators, musicians, game designers, and content creators building businesses on AI output.

The law is still in flux. Several cases are winding through U.S. courts, and Congress has held hearings specifically on AI and copyright. The European Union's AI Act and its copyright directive approach the problem somewhat differently. We'll examine those differences in Lesson 3.

Key Principle

The more human creative decision-making you can document and demonstrate — prompts, iterations, edits, selections, arrangements — the stronger your potential copyright claim on AI-assisted work. Document your process.

CopyrightA legal right granting creators exclusive control over reproduction, distribution, and derivative works for a fixed term — requiring human authorship under current U.S. law.
Public DomainWorks not protected by copyright (expired, forfeited, or never qualifying) that anyone may freely use without permission.
Human Authorship RequirementThe Copyright Office's position that copyright cannot vest in works created autonomously by AI, machines, or non-human agents.

Quiz — Lesson 1

Copyright Was Not Built for This
In the Zarya of the Dawn case, what did the U.S. Copyright Office protect?
Correct. The Copyright Office split the registration: Kashtanova's human creative choices (text, selection, arrangement) were protected; the AI-generated images themselves were not, because they lacked human authorship.
Not quite. The office made a nuanced split decision — protecting only the elements reflecting human creative choice, not the raw AI-generated images.
Which Supreme Court case first established that photographs could be copyrighted based on the photographer's creative choices?
Correct. The 1884 Burrow-Giles case established that a photograph of Oscar Wilde was copyrightable because the photographer made expressive choices about lighting, pose, and composition.
That's not it. Burrow-Giles Lithographic Co. v. Sarony (1884) was the foundational photography copyright case. Feist is about facts and databases; Thaler is the recent AI case.
Under current U.S. Copyright Office guidance, which scenario is MOST likely to yield a valid copyright for the human creator?
Correct. The selection, arrangement, and original commentary are human creative acts. This is the same logic the Copyright Office used to partially protect Zarya of the Dawn.
Not quite. Short prompts yielding verbatim AI outputs are the hardest to protect. The scenario involving selection, arrangement, and original written commentary gives the strongest human-authorship argument.
What happens to an AI-generated work that the Copyright Office determines lacks human authorship?
Correct. Without a qualifying human author, the work has no copyright owner and goes immediately into the public domain — free for anyone to copy, sell, or remix.
Not correct. Copyright doesn't transfer to the AI company by default. Works lacking human authorship go directly into the public domain with no owner at all.

Lab 1 — The Authorship Test

Practice applying the human authorship requirement to real creative scenarios

Your Challenge

You're going to present creative scenarios to the AI lab assistant and ask it to evaluate whether the human creator likely has a copyright claim — and why. Push back, ask follow-up questions, and test the edges of the rule.

Try: "A photographer uses AI to remove all the backgrounds from 200 of her photos and replace them. Does she still have copyright on the final images?" — or invent your own scenario.
AI Lab Assistant
Copyright & Authorship
Welcome to Lab 1. I'm here to help you work through the human authorship requirement in copyright law. Give me a creative scenario — something involving AI-generated or AI-assisted work — and I'll help you reason through whether the human creator likely holds a valid copyright claim. There are no easy answers here; this is genuinely contested legal territory.
Lesson 2 · Module 2

What the AI Companies Actually Own

The terms of service you agreed to — and what they say about your prompts, your outputs, and your data.
When you create something using an AI tool, who is the silent co-owner?

In 2022, when Stability AI launched Stable Diffusion as open-source software, users were surprised to discover it had been trained on billions of images scraped from the internet — including works by living artists who had never consented. Artists Sarah Andersen, Kelly McKernan, and Karla Ortiz filed a class-action lawsuit in January 2023 against Stability AI, Midjourney, and DeviantArt, alleging copyright infringement in the training data itself. The case, Andersen v. Stability AI, is still moving through Northern District of California courts.

Meanwhile, the platforms these tools run on were quietly writing contracts that shaped who owned everything created on them. Most users clicked "I agree" without reading a word.

A Tour of Real Terms of Service

Every major AI creative platform has terms of service (ToS) governing what happens to your inputs (prompts, uploaded files, reference images) and outputs (the generated content). These terms vary significantly — and they change frequently.

OpenAI
Current ToS (as of 2024): OpenAI assigns ownership of outputs to you "to the extent permitted by law." This is a significant caveat — it doesn't override the Copyright Office's human authorship requirement. OpenAI retains a broad license to use your inputs to improve its models unless you opt out via API settings. Commercial use of outputs is permitted for paid subscribers.
Midjourney
Current ToS: Free users grant Midjourney a perpetual, worldwide, royalty-free license to use, reproduce, and display their prompts and outputs. Paid subscribers retain more rights but still grant Midjourney a broad license. Only paid subscribers ($10+/month) can use outputs commercially — and even then, Midjourney claims the right to use those images in its marketing.
Adobe Firefly
Design choice: Adobe trained Firefly exclusively on licensed Adobe Stock images and public domain content — specifically to give commercial users cleaner IP indemnification. Adobe Enterprise customers receive explicit IP indemnification for Firefly outputs. This is a direct competitive response to the legal uncertainty surrounding other tools.
Stability AI
Stable Diffusion open-source: Because the model weights are open-source, the ToS is essentially the CreativeML Open RAIL-M license — which permits commercial use but prohibits generating illegal content. The lack of centralized control makes enforcement extremely difficult.
Key Case — Getty Images v. Stability AI (2023)

Getty Images filed suit in both the U.S. and UK, alleging Stability AI scraped and trained on over 12 million Getty images without a license, including images with Getty watermarks visible in some AI outputs. The UK case is proceeding; in the U.S., Judge Stephanos Bibas allowed the copyright infringement claims to proceed in August 2023. This is the first major case to directly test whether training on copyrighted images constitutes infringement.

Your Inputs May Not Be Private

Many AI platforms treat your prompts and uploaded content as training data for future model versions. In April 2023, Samsung engineers accidentally leaked proprietary chip design data through ChatGPT when they used it to debug code — the conversations were stored and potentially visible to OpenAI staff. Samsung subsequently banned ChatGPT for internal use.

OpenAI and other providers typically offer opt-out mechanisms for enterprise users, but consumer accounts often default to opt-in for training data use. Practically: do not put trade secrets, client data, or personal information into free AI tools.

The Indemnification Question

When you use AI-generated output commercially and a third party sues claiming the output infringes their copyright, who covers your legal costs? Most platforms offer no indemnification to free users. Adobe and Microsoft (Copilot) have made commercial indemnification explicit selling points for enterprise clients.

In February 2024, Microsoft launched its "Copilot Copyright Commitment" — promising to defend customers sued for copyright infringement from Copilot outputs, provided the customers used the built-in guardrails. Adobe made a similar commitment for Firefly Enterprise. These commitments do not apply to consumer-tier products.

Practical Guidance

Before using any AI tool commercially: (1) Read the current ToS — they change. (2) Check whether commercial use requires a paid tier. (3) Understand what license you grant to the platform. (4) For high-stakes commercial work, consider platforms offering explicit IP indemnification (Adobe Firefly Enterprise, Microsoft Copilot).

Terms of ServiceA contract between a platform and its users governing rights to inputs, outputs, and data — legally binding when you click "I agree."
IP IndemnificationA contractual promise by a platform to cover a user's legal costs if a third party sues over copyright infringement in the platform's AI outputs.
Training Data LicenseThe permission (or lack thereof) authorizing an AI company to use your content to improve its model — a major source of ongoing litigation.

Quiz — Lesson 2

What the AI Companies Actually Own
Why did Samsung ban ChatGPT for internal use in 2023?
Correct. Samsung engineers used ChatGPT to debug code and troubleshoot chip designs, inadvertently sharing proprietary information that was stored by OpenAI. Samsung subsequently banned the tool for internal business use.
Not quite. The Samsung incident was a data privacy issue — engineers put proprietary chip design information into ChatGPT, which stored it. This is the canonical example of why sensitive data shouldn't go into consumer AI tools.
What distinguishes Adobe Firefly from Midjourney and DALL-E in terms of commercial IP risk?
Correct. Adobe deliberately trained Firefly on licensed Adobe Stock and public domain material specifically to reduce infringement risk and offer enterprise customers explicit IP indemnification — a direct response to legal uncertainty around competitors.
Not quite. The key distinction is Adobe's training data choice (licensed/public domain only) and its IP indemnification commitment for enterprise users — designed to make Firefly safer for commercial work than tools trained on scraped internet data.
In Getty Images v. Stability AI, what was the core allegation?
Correct. Getty alleged Stability AI used 12+ million licensed Getty images as training data without authorization — with some AI outputs even displaying degraded Getty watermarks, suggesting direct reproduction.
Not quite. The core allegation was unauthorized use of Getty's images as training data — one of the first major cases to test whether training on copyrighted material constitutes infringement.
What is Microsoft's "Copilot Copyright Commitment"?
Correct. Launched in February 2024, Microsoft's commitment means enterprise customers using Copilot who face infringement suits over AI-generated content can count on Microsoft to cover legal defense — if they used the product's content filters as designed.
Not quite. The Copilot Copyright Commitment is an indemnification promise — Microsoft pays enterprise customers' legal defense costs in infringement suits over Copilot outputs, subject to conditions. It doesn't guarantee outputs are infringement-free.

Lab 2 — Reading the Fine Print

Analyze terms of service scenarios with an AI legal reasoning assistant

Your Challenge

You're a freelance designer considering using AI tools for client work. Ask the assistant to help you evaluate which tools are safer for commercial use, what questions to ask platforms, and how to protect yourself contractually when delivering AI-assisted work to clients.

Try: "My client wants AI-generated illustrations for their product packaging. Which AI tool has the least IP risk and what should I disclose?" — or ask about a real scenario you face.
AI Lab Assistant
Terms of Service & IP Risk
Welcome to Lab 2. I'm here to help you navigate the IP and contractual landscape of commercial AI tool use. Tell me about a real or hypothetical scenario — a client project, a platform you're evaluating, a contract question — and we'll work through the risks and protections together. Note: I can help you reason through issues, but this isn't legal advice for your specific situation.
Lesson 3 · Module 2

Training Data and the Art of Borrowing

AI systems learn from human work. The question of whether that learning is legal — and fair — is the central IP battle of our time.
Is training an AI on someone's work fundamentally different from a human artist studying it?

Three visual artists — Sarah Andersen, Kelly McKernan, and Karla Ortiz — filed suit against Stability AI, Midjourney, and DeviantArt. Their complaint ran to forty-four pages. The central argument: their life's work had been scraped from the internet without consent, compressed into model weights, and was now being commercially exploited to generate images in their styles — replacing demand for their actual labor.

In the same month, Getty Images filed separately, noting that some Stable Diffusion outputs displayed distorted versions of Getty's watermark — apparent evidence that images had been reproduced so faithfully that even the watermark transferred.

What "Training" Actually Means

When an AI company says it "trained" on a dataset, it means the model processed millions or billions of examples to learn statistical patterns — which pixel colors follow which, which words follow which, which musical notes follow which. The question before the courts is: does ingesting a copyrighted work to extract statistical patterns constitute the kind of "copying" copyright law prohibits?

There are two serious legal theories in play:

Fair Use (for AI training): U.S. copyright law permits "fair use" of copyrighted material in certain circumstances. Courts weigh four factors: (1) the purpose and character of the use — is it transformative? (2) the nature of the copyrighted work, (3) the amount used, and (4) the effect on the market for the original. AI companies argue training is "highly transformative" — no specific work is reproduced, only patterns are learned. Critics argue the fourth factor weighs heavily against AI: if AI output can substitute for original human work, it harms the market.

The "memorization" problem: Research by researchers at Google DeepMind and elsewhere has shown that large language and image models can sometimes reproduce training data nearly verbatim when prompted in specific ways. This undercuts the "no reproduction occurs" argument.

Documented Case — NYT v. OpenAI (December 2023)

The New York Times filed suit against OpenAI and Microsoft alleging that GPT-4 was trained on millions of Times articles without authorization. The complaint included striking examples where GPT-4 could reproduce verbatim passages from Times journalism when prompted. The Times argued this both infringes copyright and substitutes for Times content (the fourth fair use factor). The case is ongoing and widely considered the most consequential AI copyright case in U.S. history.

The Style Problem

Copyright does not protect artistic style — only specific expression. You cannot copyright "impressionist painting" or "cyberpunk aesthetic." However, when AI systems can generate images "in the style of [living artist]" with enough fidelity to suppress market demand for that artist, many creators argue the legal framework fails to provide adequate protection even when technically correct.

In August 2023, the Authors Guild — representing over 13,000 authors — sent a letter to major AI companies asking them to seek permission and pay licensing fees for training data. Signatories included Margaret Atwood, Roxane Gay, and Dan Brown. The Authors Guild filed a class action in September 2023, alleging OpenAI trained on hundreds of thousands of copyrighted books without authorization.

Opt-Out Systems and Their Limits

Several companies have created opt-out mechanisms for creators:

Google: Added an opt-out meta tag standard (robots.txt-style) that websites can use to signal they don't want their content used for AI training. Critics note this puts the burden on creators, not scrapers.

DeviantArt: Added an "noai" tag system to let artists flag work as off-limits for training. Effectiveness depends entirely on whether AI companies honor the tags.

Spawning's "Have I Been Trained?" tool: Lets artists check if their work appeared in LAION-5B (a major training dataset) and submit opt-out requests. As of 2024, Stability AI has committed to honoring these requests for future model versions.

The fundamental tension: opt-out systems mean consent is assumed unless you act, which inverts most people's intuitions about property rights. The EU's approach in the AI Act and Digital Single Market Directive requires opt-in licensing for certain AI training use — a significantly different baseline.

The U.S. Approach (Current)

Opt-out default: scraping is presumed acceptable unless content owner objects. Fair use doctrine may protect training. Courts still deciding. Burden on creators to protect themselves.

The EU Approach (DSM Directive)

Article 4 permits text and data mining unless rights holders have "reserved" their rights. Research and cultural institutions have broader TDM rights. More creator-protective but complex in practice.

What This Means for Creators

If you are a creator: use available opt-out mechanisms (Spawning, noai tags, robots.txt). If you are an AI user building commercial products: understand that the tools you use may be entangled in litigation whose outcomes could affect your rights to those outputs retroactively.

Fair UseA U.S. copyright doctrine permitting limited use of copyrighted material without permission, analyzed under a four-factor test. Whether AI training qualifies is actively contested in courts.
Text and Data Mining (TDM)The automated analysis of large datasets to extract patterns — the technical process underlying AI training. EU and U.S. law treat TDM licensing very differently.
MemorizationThe documented phenomenon where AI models can reproduce near-verbatim excerpts from their training data — relevant to copyright infringement claims because it shows actual reproduction occurs.

Quiz — Lesson 3

Training Data and the Art of Borrowing
Which of the four fair use factors do critics argue weighs most heavily AGAINST AI training on copyrighted work?
Correct. Critics argue that AI outputs can substitute for the original creator's work — an illustrator's commissions dry up because clients use AI "in their style" instead. This market substitution effect is the strongest argument against fair use for AI training.
Not quite. While all four factors matter, critics most heavily emphasize the fourth factor — market effect. If AI-generated work substitutes for the original creator's market, that weighs significantly against fair use.
What made the NYT v. OpenAI lawsuit unusually compelling as a legal case?
Correct. The Times' complaint included striking examples where GPT-4 would reproduce near-verbatim paragraphs of Times journalism when prompted — direct evidence of memorization and reproduction, which undercuts the "no actual copying occurs in training" argument.
Not quite. The legally significant element was verbatim reproduction examples in the complaint — showing that GPT-4 had memorized and could reproduce actual Times content, not just learned its general style.
What is Spawning's "Have I Been Trained?" tool designed to do?
Correct. Spawning built the tool to give artists visibility into whether their work ended up in LAION-5B — one of the major image training datasets — and to submit opt-out requests that some AI companies (including Stability AI) have committed to honoring.
That's not it. Spawning's tool addresses creator opt-out rights — specifically, checking for inclusion in LAION-5B and requesting removal from future training runs.
How does the EU's Digital Single Market Directive differ from the U.S. default on AI training?
Correct. The EU's DSM Directive Article 4 creates a TDM exception but allows rights holders to "reserve" their rights against it — which shifts the baseline compared to the U.S., where fair use is the main argument and the burden is on creators to opt out.
Not quite. The key difference is the opt-in vs. opt-out default. The EU's framework under the DSM Directive is more creator-protective, requiring rights holders to affirmatively reserve rights rather than having to opt out of assumed scraping permission.

Lab 3 — The Training Data Debate

Reason through the competing arguments in AI training data law

Your Challenge

You're going to steelman both sides of the training data debate. First, build the strongest possible argument that AI training on copyrighted work is fair use. Then, build the strongest case against it. The assistant will push back and deepen your reasoning.

Try: "Make the strongest argument that AI training on copyrighted works is fair use." Then: "Now give me the best counterargument." — or bring your own angle.
AI Lab Assistant
Training Data & Fair Use
Welcome to Lab 3. We're exploring one of the most contested questions in IP law right now: is training an AI on copyrighted works fair use? I'll help you build and stress-test arguments on both sides. This isn't just academic — courts in the U.S. and EU are deciding this right now, and the outcomes will shape the entire AI creative industry. Where would you like to start?
Lesson 4 · Module 2

Protecting Your Work in the AI Era

Practical strategies for creators who want to assert ownership, document authorship, and build defensible rights in AI-assisted work.
Given all this uncertainty, how do you actually protect what you make?

After the Zarya of the Dawn decision, the Copyright Office published a broader statement of policy: AI-generated content with "sufficient human authorship" could be registered, but applicants must disclose the use of AI and describe the human creative contribution. The office was deliberately not drawing a bright line — it would evaluate cases individually.

This created both an opening and a strategy: document the human creative choices. The more a creator could point to specific decisions — which of 300 outputs to select, how they were arranged, what was edited, what original elements were added — the stronger the registration case.

The Documentation Strategy

The single most valuable thing you can do to protect AI-assisted creative work is document your process in real time. This means:

Save prompt iterations: Keep a record of the prompts you tried, in order. A hundred-word highly specific prompt demonstrates more creative input than a five-word one. Screenshot or export your conversation history.

Keep rejected outputs: If you generated 200 images and chose 12 for your book, the selection process itself is a creative act. Save evidence you made those choices. A folder of 200 images with 12 flagged is documentation.

Record your edits: Every Photoshop layer, every text edit, every element you added on top of the raw AI output is evidence of human creative contribution. Save layered files, not just flattened outputs.

Write a process note: A short written description of how you created the work — what AI tools you used, what you decided, what you changed — is useful both for Copyright Office registration and for defending against infringement claims.

Copyright Registration in Practice

The Copyright Office has begun processing registrations for AI-assisted works. As of 2024, applicants must: (1) disclose which elements were AI-generated, (2) describe the human authorship being claimed, and (3) disclaim the AI-generated portions. This creates a patchwork registration covering only the human-created elements. Registration is still highly recommended — it establishes a public record and is required to sue for infringement in U.S. federal court.

Watermarking and Provenance

Several technical standards are emerging for proving the provenance of creative work:

C2PA (Coalition for Content Provenance and Authenticity): An open technical standard backed by Adobe, Microsoft, Google, Intel, and others. C2PA embeds cryptographically signed metadata into image, audio, and video files recording when they were created, with what tools, and by whom. Adobe's Content Credentials system implements C2PA — when you export from Photoshop with Content Credentials enabled, a verifiable record is attached.

OpenAI's DALL-E 3 watermarking: DALL-E 3 images include metadata indicating AI generation. As of 2024, OpenAI has begun testing invisible watermarking in generated images. These can potentially survive basic editing operations.

The limitation: Metadata can be stripped. Any JPEG saved through a basic converter loses embedded metadata. Watermarking is a provenance tool, not an enforcement tool — it helps you prove what you made, not prevent others from copying it.

Contractual Protections

When delivering AI-assisted work to clients, your contract matters more than any copyright claim. Consider:

Disclosure clauses: Many clients now require disclosure of AI use. Failing to disclose when asked may constitute misrepresentation. Build disclosure into your client agreements proactively.

Warranty limitations: Rather than warranting that your work is 100% original (difficult to guarantee with AI), warrant that you have the rights to use the tools you used, that you've complied with applicable ToS, and that you're not aware of specific third-party claims. This is more honest and legally defensible.

Indemnification caps: If a client insists on IP indemnification for AI-assisted work, negotiate a cap tied to the project fee rather than open-ended indemnification. The legal exposure on AI work is genuinely uncertain.

Looking Ahead: Legislative and Industry Developments

The legal landscape is changing fast. As of mid-2025, the following are in play:

2023
Multiple class actions filed against Stability AI, Midjourney, OpenAI, and others. Congressional hearings held on AI and copyright. Copyright Office launched a formal study on AI and copyright law.
2024
Copyright Office released AI Policy Report (Part 2) in July 2024 — concluding that current copyright law is generally adequate but recommending Congress consider whether additional protection for human creators harmed by AI substitution is warranted.
2024–25
Licensing deals emerging: OpenAI signed content licensing deals with the Associated Press, Axel Springer, News Corp, and others. These agreements suggest a market-based solution may develop alongside litigation — creators licensing training rights rather than relying solely on opt-out.
EU 2024
EU AI Act entered into force August 2024. General-purpose AI models must publish "sufficiently detailed summaries" of training data. Providers must comply with copyright law, including opt-outs under the DSM Directive. Enforcement begins 2025.
Your Takeaway

Perfect protection doesn't exist yet, but imperfect protection is still valuable. Register what you can. Document your process. Choose your tools carefully. Use contracts wisely. Monitor developments — this area of law is moving faster than any other in the creative industries, and the rules that apply today may change significantly within a year.

C2PACoalition for Content Provenance and Authenticity — an open technical standard embedding cryptographically signed creation metadata into media files, implemented by Adobe, Microsoft, Google, and others.
Content CredentialsAdobe's implementation of C2PA, attaching verifiable creation records to files exported from Adobe applications.
Process DocumentationThe practice of preserving evidence of human creative choices — prompts, rejected outputs, edit histories — to support copyright registration and infringement defense for AI-assisted work.

Quiz — Lesson 4

Protecting Your Work in the AI Era
What must applicants disclose when registering AI-assisted works with the U.S. Copyright Office as of 2024?
Correct. The Copyright Office now requires applicants to disclose AI-generated elements, describe the specifically human creative contributions being claimed, and disclaim the AI-generated portions — creating a patchwork registration covering only the human elements.
Not quite. The Copyright Office requires three things: identifying AI-generated elements, describing the human authorship being claimed, and formally disclaiming the AI portions. Failing to disclose AI use can invalidate a registration.
What is the C2PA standard and who backs it?
Correct. C2PA (Coalition for Content Provenance and Authenticity) is an industry-led open standard for embedding verifiable creation records in files. Adobe's Content Credentials system is the most prominent implementation.
Not quite. C2PA is an industry-led open standard (not a government regulation) that cryptographically signs creation metadata into media files — allowing anyone to verify where and how a file was created. Adobe's Content Credentials is a major implementation.
OpenAI signed training data licensing deals with which major news organizations by 2024–25?
Correct. OpenAI struck deals with the AP, Axel Springer (Politico, Business Insider), and News Corp (Wall Street Journal, NY Post) — suggesting that licensing may emerge as a market-based complement to litigation for resolving training data rights.
Not quite. The documented licensing deals were with the Associated Press, Axel Springer, and News Corp — notably not the New York Times, which instead filed a major lawsuit against OpenAI in December 2023.
What is the critical limitation of metadata-based watermarking for creator protection?
Correct. Embedded metadata is fragile — a simple resave through any basic image converter typically strips it. C2PA and Content Credentials help you prove what you made, but they don't prevent copying or enforce your rights automatically.
Not quite. The practical limitation is fragility: metadata can be trivially stripped by resaving files. Watermarking is a provenance record, not an enforcement shield — it helps you prove origin but doesn't prevent infringement.

Lab 4 — Your Protection Strategy

Build a real protection plan for AI-assisted creative work

Your Challenge

You've been commissioned to create a series of AI-assisted illustrations for a published book. Work with the assistant to build a complete protection strategy — documentation, copyright registration approach, contract terms, and disclosure practices. Make it specific to your actual situation if you have one.

Try: "I'm illustrating a children's book using Midjourney plus heavy Photoshop editing. What should my copyright strategy look like?" — or describe your real project.
AI Lab Assistant
Creator Protection Strategy
Welcome to Lab 4. We're building practical protection strategies for AI-assisted creative work. Tell me about a real or hypothetical project — the tools you'd use, the type of output, and who the client or audience is — and I'll help you think through documentation practices, copyright registration, contract terms, and disclosure. Let's make this concrete and actionable.

Module Test — Who Owns What AI Makes?

15 questions · Pass at 80% (12/15) · All lessons covered
1. The U.S. Copyright Office's position on purely AI-generated images (no human creative choices) is that they are:
Correct. Without human authorship, there is no copyright, and the work goes directly into the public domain.
Not correct. The Copyright Office holds that purely AI-generated work lacks the human authorship required for copyright and enters the public domain immediately.
2. In Thaler v. Perlmutter, what did Judge Howell rule?
Correct. Judge Howell's exact words: "human authorship is a bedrock requirement of copyright." She upheld the Copyright Office's refusal to register the DABUS painting.
Not quite. Judge Howell upheld the Copyright Office's refusal and stated that human authorship is a fundamental copyright requirement — not something that can be delegated to or held by a machine.
3. Which scenario best illustrates the kind of human creative contribution that the Copyright Office has indicated supports copyright registration for AI-assisted work?
Correct. Selection from many outputs, substantial editing, arrangement, and original framing text all constitute human creative choices — precisely the approach the Copyright Office indicated can support registration.
Not quite. The scenario with the most human creative contribution — selection from many options, significant editing, original arrangement — is the most protectable under current Copyright Office guidance.
4. Midjourney's free tier terms of service require that users grant Midjourney:
Correct. Midjourney's free tier grants very broad rights — perpetual, worldwide, royalty-free — to use your prompts and outputs. Paid subscribers retain more rights but still grant a broad license.
Not quite. Midjourney's free tier terms grant a perpetual, worldwide, royalty-free license for the platform to use your prompts and outputs — including for marketing purposes.
5. What was the Samsung ChatGPT incident primarily an example of?
Correct. Samsung engineers put proprietary semiconductor design information into ChatGPT during debugging — data that was retained by OpenAI. This became a canonical example of input data privacy risks in AI platforms.
Not quite. Samsung's ban resulted from engineers accidentally putting proprietary chip design data into ChatGPT — which stored it. The risk was data privacy, not copyright or hacking.
6. The four-factor fair use test in U.S. copyright law asks courts to consider all of the following EXCEPT:
Correct. The four fair use factors are: (1) purpose and character, (2) nature of the work, (3) amount used, and (4) market effect. Author nationality is not a factor.
Not quite. Author nationality has no role in the fair use analysis. The four factors are purpose/character of use, nature of the original, amount used, and effect on the market.
7. The "memorization" problem in AI models is legally significant because:
Correct. Research showing AI models can reproduce training data nearly verbatim undercuts AI companies' argument that training is purely transformative and involves no copying — central to the fair use defense.
Not quite. The legal significance is that memorization proves actual reproduction can occur during inference — directly challenging the AI industry's argument that training is purely transformative with no copying.
8. Which platform explicitly trained its AI image generator on licensed and public domain content to reduce IP infringement risk for commercial users?
Correct. Adobe trained Firefly exclusively on licensed Adobe Stock images and public domain content, specifically to give enterprise users cleaner IP rights and offer indemnification — a deliberate competitive differentiator.
Not quite. Adobe Firefly was the platform that deliberately trained only on licensed and public domain content, enabling the IP indemnification commitment for enterprise customers.
9. The Artists' class action (Andersen v. Stability AI) was filed in January 2023. The three named plaintiffs were:
Correct. Artists Sarah Andersen, Kelly McKernan, and Karla Ortiz were the named plaintiffs in the January 2023 class action against Stability AI, Midjourney, and DeviantArt.
Not quite. The three artist plaintiffs were Sarah Andersen, Kelly McKernan, and Karla Ortiz — visual artists who argued their styles were being exploited without consent or compensation.
10. Under the EU's Digital Single Market Directive, AI companies can train on copyrighted works unless:
Correct. Article 4 of the DSM Directive creates a text and data mining exception but allows rights holders to opt out by "reserving" their rights — shifting the default compared to the U.S. opt-out system.
Not quite. The DSM Directive's TDM exception applies unless rights holders actively reserve their rights — a different baseline from the U.S., where scraping is broadly presumed acceptable unless objected to.
11. What does the Copyright Office require creators to do when registering AI-assisted works?
Correct. The Copyright Office's current process for AI-assisted works requires disclosure of AI elements, description of the specifically human creative contributions, and formal disclaimer of the AI-generated portions.
Not quite. The three required steps are: identify which elements are AI-generated, describe what human authorship is being claimed, and formally disclaim the AI portions. No extra fee or waiting period is required.
12. C2PA content credentials are primarily useful for creators because:
Correct. C2PA creates cryptographically signed provenance records — helping creators prove what they made and how. They are a documentation tool, not a scraping prevention or copyright transfer mechanism.
Not quite. C2PA credentials establish verifiable provenance — a record of creation that can support claims about who made something and how. They don't prevent scraping or change copyright ownership.
13. Microsoft's Copilot Copyright Commitment applies to:
Correct. The Commitment protects enterprise customers (not free consumer users) who use Copilot's guardrails as designed — Microsoft won't extend the indemnification if users deliberately bypass the content filters.
Not quite. The Copilot Copyright Commitment covers enterprise customers only, and only when they use the product's built-in guardrails. It doesn't extend to free consumer accounts.
14. The NYT v. OpenAI case is particularly significant because the complaint included:
Correct. The verbatim reproduction examples in the complaint were legally powerful — direct evidence of memorization, showing that actual copying (not just pattern learning) occurs in AI training and inference.
Not quite. The complaint's most powerful evidence was verbatim reproduction — GPT-4 producing near-exact passages of Times journalism when prompted, demonstrating that training involved actual copying and not just abstract pattern learning.
15. When delivering AI-assisted creative work to clients, the BEST contract approach to managing IP indemnification risk is:
Correct. Proactive disclosure, honest warranty framing (ToS compliance rather than absolute originality), and capped indemnification reflect realistic risk allocation for AI-assisted work — legally defensible and commercially honest.
Not quite. The most defensible approach is to disclose AI use, warranty what you can honestly verify (ToS compliance), and cap indemnification at the project fee — reflecting the genuine legal uncertainty rather than making warranties you can't keep.