A note about this module: "Protecting Your Creative Voice" is designed specifically with young creators in mind — artists, writers, musicians, and makers who are just beginning to build their creative identity. The real-world cases here involve professional creators, but the lessons apply directly to you, whether you're posting art on Instagram, writing fanfiction, composing beats, or building a portfolio for college applications. Your voice matters now — not someday.
In September 2023, visual artist Hollie Mengert discovered that her distinctive cartoon illustration style — developed over a decade working for Disney and Dreamworks — had been used to train an AI image generator called Hollie+ on the platform Invoke AI, without her knowledge or consent. The model was specifically named after her. Anyone could type her name and generate unlimited images in her recognizable style. She had not been asked. She had not been paid. She had not even been told.
Mengert's response on social media sparked a wider conversation: what exactly had been taken? Not individual drawings — those remained hers. What had been extracted was something harder to name: the accumulated pattern of her choices, her color preferences, her way of drawing eyes and hands and expressions. Her creative voice, reduced to a mathematical model.
Creative voice is not the same as skill. Skill is learnable — you can practice perspective drawing, chord progressions, or sentence structure until you're technically proficient. Voice is something different: it's the constellation of choices you make consistently that no one taught you to make that way.
A writer's voice includes their sentence rhythm, the metaphors they reach for, the details they notice, and what they leave out. A visual artist's voice includes their linework, their color instincts, their compositional habits. A musician's voice includes their tonal preferences, their timing idiosyncrasies, their harmonic comfort zones.
Voice develops slowly, often over years. It is frequently the thing young creators are least aware of — you often don't know what makes your work distinctively yours until someone points it out, or until you see an AI imitate it.
AI image and text generators work by finding statistical patterns in large datasets of human-made work. When trained on a specific artist's output, a model can learn which visual patterns co-occur — certain line weights alongside certain color palettes, certain compositional choices alongside certain subject matter. The output can be eerily convincing.
But approximation is not identity. The model has no reason for the choices it makes — no lived experience that led to preferring that particular shade of blue, no memory of the teacher who showed them that brushstroke technique, no emotional state driving the compositional tension. The output may look like Hollie Mengert. It is not Hollie Mengert thinking.
This distinction matters practically: an AI-generated image in your style lacks your intentionality. It cannot make the choice to break from your typical patterns for a specific emotional reason. It will reproduce your habits endlessly but cannot develop them. Your creative voice is not static — it grows, responds, rebels against itself. A model trained on you today captures who you were, not who you are becoming.
Your voice is still forming — and that is not a weakness. It means no snapshot of your current work fully captures you. The thing AI cannot steal is your next evolution. Keep creating, keep changing, keep surprising yourself.
Here is the uncomfortable legal reality: in the United States and most other countries, style itself cannot be copyrighted. Copyright protects specific creative works — a particular drawing, a specific song recording, a written text. It does not protect the general approach, look, or feel that makes those works recognizable.
This means that if an AI generates an image "in the style of" your work without copying any specific piece of your art, there is currently no clear legal remedy. The Hollie Mengert situation exposed this gap vividly: the AI company had technically scraped publicly available images (a practice that remains legally contested) and produced outputs that weren't copies of any individual work.
Several lawsuits filed in 2023 — including one by a group of artists against Stability AI, Midjourney, and DeviantArt — are attempting to establish new legal frameworks. As of this writing, these cases have not produced definitive rulings. The law has not caught up with the technology.
Because the law does not yet protect creative style, understanding how to protect and assert your voice through non-legal means becomes especially important for working creators — and for students building their creative identity now.
In this lab, you'll work with the AI to examine your own creative work — whether you're a writer, visual artist, musician, or any other kind of maker — and identify the patterns that might constitute your emerging creative voice. The AI will ask you questions and help you build a "voice map."
Complete at least 3 exchanges to finish this lab.
In January 2023, a dataset called LAION-5B — containing roughly 5.85 billion image-text pairs scraped from the public internet — was revealed to be the primary training source for Stable Diffusion, DALL-E 2, and Midjourney. Among those images were millions of pieces of art from platforms like DeviantArt, ArtStation, and personal artist websites.
The website haveibeentrained.com was launched, allowing artists to search whether their work was included. Thousands of professional artists discovered their portfolios had been scraped wholesale. Many had posted their work publicly, but with no expectation — and no disclosed policy — that it would be used to train commercial AI systems. The Terms of Service on most platforms did not clearly address this use case when the scraping occurred.
Web scraping is the automated collection of publicly accessible online content. Programs called "crawlers" or "spiders" systematically visit URLs, download content, and store it. This practice has existed for decades — it's how search engines index the web. What changed is the scale and the commercial purpose: instead of indexing content for search, AI companies are using scraped content to train generative models.
When you post an image to a public Instagram account, ArtStation portfolio, or personal website, that image is technically accessible to any web crawler. Unless the platform has specific technical protections (robots.txt directives, rate limiting, login walls) or contractual restrictions, there is nothing technically preventing scraping.
Critically: most of this scraping happened before these practices were widely understood or disclosed. Artists who posted work in 2015 or 2018 had no way of knowing their images would be used to train a text-to-image model in 2022.
Following widespread backlash from the artist community, several platforms and AI companies introduced opt-out mechanisms. ArtStation added a "Do Not Train" tag that users can apply to their work. DeviantArt introduced a similar feature. Adobe stated that works in Adobe Stock (their paid platform) would not be used to train third-party models without creator consent.
However, these opt-out systems have significant limitations. They are prospective — they apply to future scraping, not to data already collected. They rely on AI companies voluntarily honoring the signals. And they require creators to actively take action, placing the burden on individuals rather than requiring consent by default.
The tool Glaze, developed by researchers at the University of Chicago and released in 2023, takes a different approach: it applies invisible perturbations to artwork images that cause AI models to misinterpret the style, essentially "cloaking" the work against style extraction while remaining visually unchanged to human viewers.
If you post creative work publicly, you can: (1) Apply "Do Not Train" tags on ArtStation and DeviantArt; (2) Consider using Glaze on high-resolution images of artwork before posting; (3) Check your platform's current Terms of Service for AI training clauses; (4) Keep original high-resolution files with metadata (artist name, date, copyright notice) embedded.
A fundamental question underlies all of this: should AI training on publicly available creative work require opt-in consent from creators, or should it be permitted unless creators actively opt out?
The current legal default in most jurisdictions favors permitted use unless blocked. The European Union's AI Act and the EU Copyright Directive's text and data mining provisions are moving toward stronger creator protections, but implementation remains in progress. In the US, the Copyright Office issued guidance in 2023 stating that AI-generated works with no human creative input are not copyrightable — but the question of training data rights remains under active legal consideration.
For young creators, understanding this debate matters not just as future professionals but as citizens. The norms being established now, through lawsuits, policy debates, and platform decisions, will shape the creative economy you enter.
"Publicly accessible" is not the same as "freely usable for any purpose." Public accessibility was designed for human audiences — for other artists to see your work, for potential clients, for personal expression. The extension of that accessibility to machine training at commercial scale is a new use that existing consent frameworks were not designed to govern.
In this lab, you'll audit your own digital creative presence with AI guidance. You'll examine which platforms you post on, what protections they offer, and build a personal action plan for better protecting your work going forward.
Complete at least 3 exchanges to finish this lab.
In 2023, the Writers Guild of America went on strike for 148 days — the longest Hollywood writers' strike since 1988. One of the central demands was protection against AI: specifically, that studios could not use AI-generated content to replace writers, and that any AI-assisted work would still require full writer credit and compensation.
The WGA's position was not that writers could never use AI tools. It was that writers should not be replaceable by AI at studio discretion. The final agreement, reached in September 2023, established that studios cannot require writers to use AI, that AI-generated material cannot be considered "literary material" under guild contracts, and that writers retain credit even when helping develop AI-assisted projects. The distinction the WGA drew — between using AI as a tool versus being replaced by AI as a professional — became a template for similar negotiations in other creative industries.
Not all AI use in creative work is the same. There is a meaningful spectrum between using AI as a tool that enhances your voice and using AI in ways that progressively displace your creative decision-making.
Consider a photographer who uses AI-powered selection tools in Photoshop to speed up background removal. The compositional choices, the subject selection, the lighting decisions, the aesthetic vision — all remain theirs. The AI is a more efficient eraser. Now consider a social media manager who prompts an AI image generator for every post visual, choosing from outputs, never making any compositional choices themselves. Over time, their visual sense may atrophy from disuse.
The difference is in where the creative judgment lives. When AI executes your vision, it's a tool. When AI generates options for you to select from, you're still exercising judgment — but passively. When AI makes choices you never review or understand, you've ceded creative authorship.
Novelist Roxane Gay stated publicly in 2023 that she would not use AI writing tools, arguing that the specific effort of finding words — even when it's difficult — is inseparable from her creative process. The resistance, the searching, the dissatisfaction with draft sentences, is where her voice lives.
Musician Holly Herndon took the opposite approach: she created her own AI model trained exclusively on her voice, called "Holly+," allowing fans to submit covers performed in her voice — but with her explicit consent and on her terms. She frames AI as a means to multiply her voice, not replace it.
Concept artist Karla Ortiz, one of the lead plaintiffs in the 2023 lawsuit against AI image generators, continued using AI tools for specific tasks (color palette exploration, rough spatial composition) while refusing tools trained on scraped artist data. She articulated a principle: AI tools trained with consent, on appropriate data, can be used. AI tools that benefited from unconsented exploitation of other artists' work are different in kind.
You don't need to adopt anyone else's position. But developing your own clear principles about when and how you'll use AI in your creative process — before you need them — protects your voice better than deciding case-by-case under pressure. Think about: what parts of your creative process do you want to keep entirely human? What tasks would AI assistance genuinely serve your vision?
For students and young creators, there is an additional consideration: you are still developing your skills. Using AI to skip the practice phases of creative development is different from a professional using AI to work more efficiently after years of building craft.
When a student writer uses AI to draft their essays, they may produce better-looking text — but they don't develop the ability to think through writing, to structure arguments, to find their voice through the resistance of the blank page. When a student artist uses AI to generate all their visual concepts, they may skip the hours of drawing from observation that build visual memory and spatial reasoning.
This is not an argument against all AI use in student creative work. It's an argument for intentionality: knowing which skills you're trying to build, and protecting the practice of those skills from automation.
The goal is not to avoid AI tools — it's to ensure that when you use them, you remain the creative author of the work. Your vision drives the output. Your judgment evaluates it. Your sensibility decides what's right. AI can execute, suggest, or accelerate — but the creative intelligence should remain yours.
In this lab, you'll draft your own personal policy for how you'll use — or not use — AI in your creative work. The AI tutor will challenge you to think through the edge cases, help you identify what you want to protect, and sharpen your reasoning.
Complete at least 3 exchanges to finish this lab.
In March 2023, the US Copyright Office registered a graphic novel called Zarya of the Dawn — then partially rescinded the registration. The author, Kristina Kashtanova, had used Midjourney to generate the images while writing the text and story herself. The Copyright Office ruled that the images could not be copyrighted because they were generated by AI, but the text and the selection and arrangement of the images could be — because those reflected human creative judgment.
This case established a significant precedent: the human creative decisions that go into directing, selecting, and arranging AI outputs can be copyrightable, even when the AI-generated elements themselves are not. It was simultaneously a limitation (AI outputs alone get no protection) and a clarification (your curation and direction of AI remain legally protectable expression).
One of the most practical things any creator can do — especially young creators building a portfolio — is establish a clear, timestamped record of their creative development. This serves multiple purposes: it demonstrates the evolution of your voice over time, provides evidence of creative authorship in disputes, and builds the kind of documented creative history that is increasingly valuable in professional contexts.
Specific documentation practices include: saving dated draft files (not just final versions), keeping sketchbooks or writing journals with dated entries, posting work-in-progress documentation publicly on platforms that timestamp content, and embedding copyright metadata in digital files (creator name, creation date, copyright notice) before posting anywhere.
Save dated drafts of every project. A folder showing 12 versions of a painting over three months is powerful evidence of human creative process.
Use photo/file metadata tools to embed your name, the date, and a copyright notice in digital files before uploading. This travels with the file.
Post work-in-progress to timestamped public platforms. These create verifiable records of when you made what — useful if your work is later misattributed.
Keep a dated creative journal — physical or digital. Notes about your influences, decisions, and struggles document the human creative intelligence behind your work.
The most durable protection for your creative voice is the voice itself. A distinctive, developed, continuously evolving creative identity is harder to replicate than an underdeveloped one. This isn't about becoming more technically skilled — it's about developing the specific constellation of aesthetic preferences and choices that make your work recognizably yours.
Paradoxically, deliberately studying your influences helps differentiate your voice. When you can articulate why you're drawn to a particular artist's use of negative space, or a particular author's sentence rhythm, you are building conscious aesthetic judgment — the capacity to make intentional choices rather than default ones. That intentionality is what AI currently cannot replicate.
Several professional artists and educators have noted that younger creators who engage with AI early as a reflection tool — prompting it to imitate their work and then identifying what's missing or wrong about the result — develop clearer self-awareness about their own voice than those who either avoid AI entirely or use it uncritically.
If you have access to an AI image or text generator, try this: describe your creative style in as much detail as you can and ask it to generate something in that style. Then look at the output critically — what did it get right? What is subtly or obviously wrong? What is impossible to capture? Your answers reveal what's most distinctive about your voice.
Individual protections matter, but the WGA strike demonstrated that collective action can establish industry-wide norms. As a young creator, building literacy about industry structures — unions, guilds, licensing frameworks, platform terms of service — prepares you to participate in the collective decisions that will shape the creative economy.
Organizations like the Authors Guild, the Graphic Artists Guild, and Human Artistry Campaign are actively working on AI and copyright policy. Following their work, even as a student, keeps you informed about the debates that will define your professional future.
The norms being set right now — in lawsuits, in platform policies, in guild negotiations, in government hearings — are not fixed. They are being made by people who organized, showed up, and argued for their interests. Young creators who understand these systems can participate in shaping them.
Your creative voice is not just a personal possession — it's your contribution to a larger creative culture. Protecting it is not defensive or backward-looking. It's the act of insisting that human creative intelligence has irreplaceable value, and that the tools we build should serve that intelligence rather than substitute for it.
In this final lab, you'll synthesize everything from this module into a personalized Creative Voice Defense Plan — specific, actionable steps you can start on this week to document, develop, and protect your creative identity in the AI age.
Complete at least 3 exchanges to finish this lab.