How Visual Artists Can Protect Their Work From AI Scraping, Copyright Theft, and Style Cloning in 2026

If you're a visual artist or digital illustrator reading this in 2026, you already know the problem. Your portfolio—years of work, your signature style, your creative identity—has likely been scraped into one or more AI training datasets without your knowledge, let alone your consent. Generative image models can now produce work that mimics your aesthetic in seconds. And the legal landscape is shifting fast enough that last year's advice may already be outdated.

We've written previously about what copyright law says when AI trains on your art and provided a practical opt-out guide for visual artists. This article is the 2026 update: a comprehensive legal toolkit covering the five things every visual artist should be doing right now—opting out of AI datasets, registering copyright for AI-assisted work, filing DMCA takedowns for AI-generated copies, tracking the Andersen v. Stability AI class action, and exploring style-licensing deals to monetize AI training partnerships.

1. AI Dataset Opt-Out Tools: What Actually Works in 2026

The first line of defense is preventing your work from being ingested into future AI training datasets. Several tools have matured since we first covered this space, and while none are perfect, they're worth using—especially in combination.

Spawning.ai and Have I Been Trained

Spawning.ai operates the Have I Been Trained platform, which lets artists search for their work in the LAION-5B dataset (the image-text pair dataset used to train Stable Diffusion and other models). Since launching in December 2022, Spawning has helped artists opt out over 78 million artworks from AI training. The opt-out is free for individual artists, and Spawning's API makes those opt-out requests available to any organization training models.

Stability AI committed to honoring Spawning opt-out requests for Stable Diffusion V3—a significant precedent. Spawning also partnered with ArtStation to ensure opt-out requests made on that platform are honored through the API, and with Shutterstock to opt out all images posted to their platform by default. If you haven't already, create an account at haveibeentrained.com and search for your work.

Glaze and Nightshade

Developed by the SAND Lab at the University of Chicago, Glaze and Nightshade take a different approach. Rather than opting out of datasets, they modify your images to interfere with AI training.

Glaze applies subtle, human-imperceptible perturbations to your artwork before you post it online. These perturbations confuse AI style-mimicry models, making it harder for them to reproduce your aesthetic. The current version, Glaze 2.2, is available for free download for Windows and supports newer GPU hardware.

Nightshade goes further: it's a data-poisoning tool that can cause AI models trained on Nightshade-processed images to produce garbled or incorrect outputs. The idea is to make unauthorized scraping of your work actively counterproductive for AI developers. While researchers have noted that these tools aren't foolproof—some attacks have been shown to bypass Glaze protections—the developers continue to update the tools in response, and using them raises the cost and complexity of scraping your work without consent.

The limits of opt-out

It's important to understand what these tools can and cannot do. Opt-out tools like Spawning only work if AI companies honor the requests—and not all do. Glaze and Nightshade protect images you process going forward, but they can't retroactively remove your work from datasets that already ingested it. And as we've explained before, copyright law doesn't currently require AI companies to seek your permission before training on publicly available works, though that's exactly what cases like Andersen v. Stability AI are testing.

The practical takeaway: use all three tools together. Register your work on Have I Been Trained to opt out of future training. Apply Glaze (or Nightshade, for higher-stakes work) before posting new art online. And keep reading—the enforcement landscape is evolving rapidly.

2. Copyright Registration Strategy for Digital and AI-Assisted Art

Copyright registration is the foundation of every legal protection available to you. Without a registration, you can't file a DMCA takedown or sue for infringement in federal court. The U.S. Copyright Office has issued significant guidance on AI and copyright that every visual artist needs to understand.

The Copyright Office's AI guidance

In March 2023, the Copyright Office issued registration guidance for works containing AI-generated materials. The core principle: only human-authored material is copyrightable. Works generated entirely by AI cannot be registered. But works that combine human authorship with AI-generated elements can be registered—if the human contribution is sufficiently creative and the applicant discloses the AI-generated portions.

In January 2025, the Office published Part 2 of its AI Report, which addresses the copyrightability of outputs created using generative AI. The report reaffirmed that copyright requires human authorship and clarified that simply prompting an AI system does not make you the author of the output. However, if you make substantial creative modifications to an AI-generated image—through editing, compositing, painting over, or otherwise transforming it—those human-authored modifications may be registrable.

In May 2025, the Office released a pre-publication version of Part 3, addressing the use of copyrighted materials in AI training. While this section focuses on the AI developer's side, it's directly relevant to artists because it analyzes whether unauthorized training constitutes fair use or infringement—a question the Andersen litigation is actively testing.

Practical registration strategy

For visual artists in 2026, here's what we recommend:

• Register your original digital art. If you created the work entirely by hand (or with traditional digital tools like Photoshop or Procreate, where you controlled every pixel), register it. This is straightforward and gives you the full protections of copyright law, including statutory damages and attorneys' fees if you register before infringement occurs (or within three months of publication).
• Register AI-assisted works carefully. If you used AI as part of your creative process—for example, generating a base image with Midjourney and then substantially editing, painting over, and compositing it into a larger work—you can register the human-authored portions. You must disclose the AI-generated content in your application and limit your claim to the human-authored material. The Copyright Office has been consistent about this: failing to disclose AI-generated content can result in cancellation of your registration.
• Register in batches. The Copyright Office allows you to register multiple works as a group for a single fee under certain conditions (e.g., group registration of unpublished works, or group registration of published photographs). This is cost-effective for digital artists producing large volumes of work.
• Register before you publish. Posting your work online constitutes publication. If you register before publication (or within three months after), you preserve your right to statutory damages—up to $150,000 per work for willful infringement—and attorneys' fees. These remedies are critical leverage in any enforcement action.

3. DMCA Takedowns for AI-Generated Copies of Your Work

The Digital Millennium Copyright Act (DMCA) provides a powerful, relatively inexpensive tool for removing infringing content from online platforms. If someone generates an AI image that copies your protected work and posts it online, a DMCA takedown notice can get it removed—often within days.

When a DMCA takedown applies to AI-generated content

Section 512 of the DMCA (17 U.S.C. § 512) provides a safe harbor for online service providers: if they receive a proper takedown notice and promptly remove the infringing material, they're not liable for the infringement. This applies to AI-generated copies just as it does to any other infringing content.

The key question is whether the AI-generated output is substantially similar to your protected work. If someone prompts an AI model with your name and the output closely resembles a specific piece you created, that may constitute infringement of your reproduction right. If the output is merely "in your style" but doesn't copy a specific work, it's more legally uncertain—style itself is not copyrightable, though the issue is being actively litigated.

How to file a DMCA takedown

A valid DMCA takedown notice must include:

1. Identification of the copyrighted work you claim has been infringed (or a representative list if there are multiple works).
2. Identification of the infringing material to be removed, with enough detail for the platform to locate it (URLs are standard).
3. Your contact information—name, address, phone number, email.
4. A statement that you have a good faith belief that the use of the material is not authorized by the copyright owner, its agent, or the law.
5. A statement that the information in the notification is accurate and that, under penalty of perjury, you are authorized to act on behalf of the copyright owner.
6. Your physical or electronic signature.

Most major platforms—Instagram, X (Twitter), DeviantArt, ArtStation, Reddit, and others—have online DMCA takedown forms. You don't need a lawyer to file one, though legal guidance helps if the platform pushes back or the infringer files a counter-notice.

What to do if a counter-notice is filed

If the person who posted the AI-generated copy files a counter-notice claiming they have the right to use the material, the platform will restore the content within 10-14 business days unless you file a federal lawsuit seeking a court order to restrain the infringement. This is where having a copyright registration becomes essential—you can't sue in federal court without one. This is also where having legal counsel becomes important; a lawyer can evaluate whether your infringement claim is strong enough to pursue in court and, if so, file the complaint within the required window.

4. Andersen v. Stability AI: What the Class Action Means for Artists

The most consequential litigation for visual artists is Andersen v. Stability AI, pending in the Northern District of California before Judge William Orrick. The case was filed in early 2023 by artists Sarah Andersen, Kelly McKernan, and Karla Ortiz against Stability AI, Midjourney, and DeviantArt, alleging that the companies ingested the artists' copyrighted works into their training datasets without permission.

Where the case stands

In October 2023, Judge Orrick largely granted the defendants' motions to dismiss but allowed the direct infringement claims to proceed and gave plaintiffs leave to amend. In August 2024, the court issued a significant order on the amended complaint that provided several key takeaways for artists:

• The "VCR analogy" doesn't work. AI companies have argued their technology is like the VCR—capable of both infringing and non-infringing uses. Judge Orrick rejected this framing, noting that unlike a VCR, generative AI models are "built to a significant extent on copyrighted works" and "necessarily invoke copies or protected elements of those works." This is a critical distinction that could shape how courts evaluate AI training infringement.
• Direct infringement and inducement claims survive. The court allowed plaintiffs' claims that Stable Diffusion itself is an infringing work (the "model theory") and that defendants induced infringement by distributing it to proceed through discovery.
• DMCA CMI claims were dismissed with prejudice. The court dismissed claims under § 1202(a) and (b) of the DMCA for removal or alteration of copyright management information. The 1202(b) claim failed because there were no allegations that any AI output was identical to the plaintiffs' works—an "identicality" requirement that some courts have applied. This means DMCA CMI claims may be difficult for artists to pursue unless they can show near-identical outputs.

As of 2026, the case is in active discovery. The outcome could establish whether unauthorized ingestion of copyrighted works into AI training datasets constitutes infringement—a question with enormous implications for every visual artist. The case is tracked on CourtListener (docket 3:23-cv-00201).

What this means for you

Even if Andersen succeeds, it won't automatically compensate individual artists. Class action settlements typically distribute modest per-artist amounts. But the case could establish legal precedent that makes it easier for individual artists to pursue their own infringement claims, and it could push AI companies toward licensing models rather than scraping. In the meantime, your best protection remains: register your copyrights, document where your work appears online, and keep records of any AI outputs that appear to copy your specific works.

5. Licensing Your Style for AI Training: A Monetization Path

Not every artist wants to fight AI. Some are exploring a different question: if AI companies want to train on my work, how can I get paid for it?

The market for ethical AI training data is growing. As legal pressure mounts and public scrutiny increases, major AI companies are shifting toward licensed data. Platforms like Wirestock have emerged as intermediaries, allowing artists to opt in to AI training in exchange for payment. According to Wirestock's CEO, over 300,000 creators have already been paid through AI data licensing deals, with payments ranging from microtransactions (cents per image for bulk data) up to $15-20 per artwork for specific commissioned content.

What a style-licensing deal looks like

A style-licensing agreement typically grants an AI company the right to train on your work in exchange for compensation. Key terms to negotiate include:

• Scope of the license. Which works are included? Is the license exclusive or non-exclusive? Does it cover only training, or also the right to generate outputs "in your style"?
• Compensation structure. Will you receive a flat fee, per-image payment, revenue share, or ongoing royalties? Will payments be one-time or recurring?
• Attribution and provenance. Will the AI company credit you in model documentation or outputs? Can you require that your name not be used as a prompt?
• Term and termination. How long does the license last? Can you revoke it? What happens to models already trained on your work if you terminate?
• Output restrictions. Can you restrict the types of outputs generated (e.g., no commercial use, no explicit content)?

Artists should also understand that licensing your work for AI training is distinct from licensing your copyright. A training license grants the AI company the right to ingest and learn from your work; it doesn't necessarily give them the right to reproduce your work or distribute copies. These are separate rights under copyright law, and your agreement should clearly distinguish between them.

Risks to consider

Style licensing is not without risk. Once an AI model has been trained on your work, the training can't be "undone"—your style may be embedded in the model permanently. If you later decide you don't want your aesthetic associated with a particular AI platform, you may have limited recourse. Additionally, some artists in your community may view licensing as legitimizing the same practices that others are fighting in court. These are personal and business decisions, not purely legal ones—but a well-drafted agreement can at least protect your legal rights and ensure you're compensated.

For artists considering this path, we recommend working with an attorney to review or negotiate any licensing agreement. The terms matter enormously, and the market is still young enough that there are few standard-form contracts. For more on how AI-generated content intersects with copyright ownership—and why the distinction between human and machine authorship matters—see our analysis of AI-generated music copyright after Suno and Udio, which covers parallel issues in the music space.

Actionable Next Steps

If you're a visual artist or digital illustrator, here's your prioritized action plan for 2026:

1. Run an audit. Search for your work on Have I Been Trained. Document what you find with screenshots and URLs—this creates a record that could be valuable in any future enforcement action.
2. Opt out. Submit opt-out requests through Spawning for any of your work found in training datasets. Apply Glaze to new work before posting it online.
3. Register your copyrights. File registrations with the U.S. Copyright Office for your most important works, especially before publishing them online. If you use AI tools in your process, disclose the AI-generated portions and limit your claim to your human-authored contributions. Registration is the gateway to every enforcement mechanism available to you.
4. Monitor for AI-generated copies. Set up Google Alerts for your artist name and image titles. Search AI image platforms periodically for outputs that closely resemble your work. If you find infringing copies, file DMCA takedown notices with the hosting platform.
5. Follow the Andersen case. The outcome of Andersen v. Stability AI could reshape your legal rights. Monitor developments through the CourtListener docket or legal news sources. If a favorable precedent emerges, it could open new enforcement pathways.
6. Consider licensing. If you're open to AI training partnerships, explore platforms like Wirestock or negotiate directly with AI companies. Have an attorney review any agreement before signing—these are novel contracts in a rapidly evolving market, and the terms you accept today could have long-term consequences.
7. Talk to a lawyer. If you believe your work has been infringed at scale, or if you're considering a licensing deal, consult with an attorney who understands both copyright law and the AI landscape. The cost of preventive legal advice is almost always lower than the cost of enforcement after the fact.

The tools and legal frameworks available to visual artists have evolved significantly since AI image generation first disrupted the creative landscape. The copyright system wasn't designed for this moment, but it still provides meaningful protections—if you know how to use them. Register your work. Opt out of unauthorized training. Document everything. And when the system falls short, push for the changes that artists deserve.