⚖️ Published June 6, 2026 — every claim in this article is sourced and verifiable. Key facts: On March 2, 2026, the U.S. Supreme Court denied certiorari in Thaler v. Perlmutter (Case No. 25-449), making it settled law that a work created entirely by AI, with no meaningful human authorship, cannot be copyrighted in the United States. The financial stakes are now enormous: Anthropic agreed to a roughly $1.5 billion settlement in Bartz v. Anthropic — the largest copyright settlement on record — and faces a separate ~$3 billion music-publishing suit filed in January 2026. Litigation trackers now count 166 active AI copyright cases across U.S., EU, and UK courts. The pivotal Sony fair-use ruling against Suno is expected in summer 2026, and Germany's GEMA v. Suno verdict was pushed from June 12 to July 31, 2026. Meanwhile, Suno raised another $400 million on June 3, 2026 while still being sued. The legal ground under every AI-using creator is shifting — and the rules that protect you are now knowable.
If you use AI to make anything you intend to own, sell, license, or defend — words, images, music, video, code — 2026 is the year the law stopped being theoretical. For three years, creators argued about AI copyright in comment threads. This year courts started answering, and the answers carry nine and ten-figure price tags. The good news is that the single most important rule is now clear, settled, and easy to act on. The harder news is that the rule has a sharp edge, and the difference between a work you fully own and a work anyone on earth can copy for free comes down to choices you make while you create — not after. This guide explains the rule, the money behind it, the cases that will define the next year, and the four habits that keep your work protected.
The One Rule the Supreme Court Just Locked In
On March 2, 2026, the Supreme Court declined to hear Stephen Thaler's appeal, leaving in place a D.C. Circuit ruling that human authorship is a bedrock requirement of U.S. copyright. Thaler had listed an AI system as the sole author of an image and asked the courts to recognise machine authorship. They refused at every level, and the Supreme Court's denial ended the road. The principle the courts affirmed is narrow but decisive: copyright protects what a human creates, and the author must be the person who created, operated, or directed the AI — never the machine itself. This does not ban AI-made content, slow down anyone's commercial use of it, or make AI tools risky to touch. It establishes one thing only — that output generated by AI with no meaningful human creative contribution is not eligible for copyright protection, and therefore belongs to no one.
- Purely AI-generated work is not copyrightable. Type a prompt, accept the output, publish it unchanged — and that text, image, song, or clip has no copyright protection. Anyone can reproduce, remix, or sell it, and you cannot stop them. A prompt alone is not authorship in the eyes of the Copyright Office.
- AI-assisted work usually is copyrightable — when the human contribution is substantial. If you use AI to draft, then meaningfully select, arrange, revise, and shape the result with your own creative judgment, the work you produce can be registered as yours. The protection attaches to your human creative choices, not to the raw AI output underneath them.
- You must disclaim the AI-generated portions when you register. The Copyright Office requires applicants to identify which parts were machine-generated and exclude them from the claim. Registering AI output as fully human-made is a false statement that can void the registration.
- The author is always a person, never the tool. The courts framed it cleanly: the rule does not penalise using AI — it requires that a human being who created, operated, or used the AI be named as the author. Use the tool freely; just make sure a human did the authoring that copyright actually protects.
The Money Just Got Real: Anthropic's $1.5 Billion Settlement
The authorship question is settled, but the far larger fight — whether training AI on copyrighted work without a licence is legal — is still being decided, and the numbers now make clear why companies are choosing to settle rather than gamble on a verdict. Anthropic agreed to pay roughly $1.5 billion to resolve Bartz v. Anthropic, a case brought by book authors over training data. That is the largest known copyright settlement in history, and it was rational: the theoretical exposure in a loss ran into the hundreds of billions. Anthropic separately faces a music-publishing lawsuit filed in January 2026 by Universal, Concord, and ABKCO seeking around $3 billion — described by trackers as the largest non-class-action copyright case in U.S. history. When a single training-data dispute can dwarf a company's revenue, settlements and licensing deals stop being a courtesy and become survival. That economic reality is quietly reshaping which AI tools are safe to build a business on.
166 Lawsuits Are Rewriting the Rules — The Ones That Matter
Litigation trackers now follow 166 active AI copyright cases across U.S. federal courts and European and UK courts. Most will never make headlines, but a handful will set the rules everyone else lives under. These are the ones a creator should actually watch in 2026.
- Sony v. Suno and v. Udio (fair use, ruling expected summer 2026) — the music industry's defining battle. Warner settled with Suno and Universal settled with Udio, but Sony refused both and is litigating the core question: is training an AI on copyrighted recordings transformative fair use, or industrial-scale infringement? A loss for the AI companies forces every music model to license or shut down. A win guts the labels' leverage and resets the market overnight.
- GEMA v. Suno (Munich Regional Court) — Germany's collecting society alleges Suno reproduced protected works. The verdict was moved from June 12 to July 31, 2026 for administrative reasons. A win for GEMA would be the first major European ruling that AI platforms need a licence to train on copyrighted music — and it would follow the same court's November 2025 finding against OpenAI over lyric memorisation.
- New York Times v. OpenAI and Microsoft — the Times alleges millions of its articles were copied to train ChatGPT and seeks billions in statutory damages. Summary judgment was scheduled for 2026 with discovery ongoing. This case probes the heart of the news-and-AI relationship and the legality of training on journalism.
- Getty Images v. Stability AI (UK High Court) — a closely watched European test where the court rejected a secondary copyright claim, a partial setback for rights holders that shows these outcomes are far from one-sided across jurisdictions.
- Disney v. Midjourney and the wave of likeness and character suits — studios are pursuing AI image and video generators over copyrighted characters, while text-to-video tools raise fresh questions about who owns a generated clip that looks like a protected work. New filings, including search and answer engines facing RAG-output claims, arrive almost weekly.
What This Means for You as a Creator
| Your Situation | Current Legal Status | What to Do Now |
|---|---|---|
| You write or design with AI assistance | Copyrightable if your human contribution is substantial | Keep your drafts, edits, and decision notes — they are your proof of authorship |
| You sell purely AI-generated images or text | Not copyrightable — anyone may copy or resell it | Add real creative direction: select, arrange, and revise so a human authored the final work |
| You build a product on AI music tools | Unsettled — Sony fair-use ruling expected summer 2026 | Prefer licensed AI-music platforms; check ownership terms before commercial release |
| You train a custom model on third-party data | Active legal gray area — billion-dollar exposure | Get an IP attorney's review before training or shipping |
| Your face or voice was deepfaked | Patchwork — federal protection pending, some state laws active | Use platform takedown tools and document the unauthorised use immediately |
The 4 Rules for Creators Using AI in 2026
- Document the human in the loop. The line between AI-assisted (yours) and AI-generated (no one's) is the degree of human creative control. Keep your prompts, your edits, and a short note on the choices you made. This is the evidence that turns AI output into a work you can register and defend.
- Read the terms of service of every AI tool you monetise. Ownership of outputs varies wildly between platforms — some assign rights to you, some retain control, and some explicitly tell users they are not the owner of what they generate. If you are building a catalogue or a product on a tool, know who actually owns the output before you depend on it.
- Never claim copyright on purely AI-generated work. Beyond being legally hollow, registering machine output as your own original creation is a false statement that can void the registration and expose you to liability. Claim what you authored; disclaim what the model generated.
- Use AI to augment, not to replace. The most legally durable and creatively strongest position in 2026 is using AI as a tool inside your process — research, drafting, iteration — while keeping the decisive creative judgment human. It produces better work and stronger IP protection at the same time.
The single highest-leverage habit any creator or business using AI can adopt this week: add one line to your workflow. For every AI-assisted project, save the prompts you used, the AI output you started from, the changes you made, and a sentence on why. It takes about two minutes per project. If you ever need to register a copyright, defend a licence, or prove human authorship in a dispute, that two-minute record is the entire difference between a protected work and one that has quietly fallen into the public domain. The creators who lose their rights in 2026 will rarely lose them in court — they will lose them by never keeping the record that proves a human was the author.
Maintaining human creative control is far easier when you can move between models instead of leaning on one. LumiChats gives creators and professionals access to Claude Opus 4.8 and Sonnet 4.6, GPT-5.5, Gemini 3.5, DeepSeek V4, and 40+ more models under a single ₹69/day pass (about $1/day) — so you can draft with one model, restructure with another, and keep your own editorial judgment at the centre of every decision. That cross-model workflow is exactly the kind of human-directed process that establishes the authorship copyright protects — and it costs less than a single-platform subscription.
01Can I copyright something I made with AI?
Yes — if your human contribution is substantial. Work where you used AI to draft and then meaningfully selected, arranged, and revised the result with your own creative judgment can be registered as yours. Work generated entirely by AI from a prompt, with no real human authorship, cannot. When you register, you must disclaim the purely AI-generated portions.
02What did the Supreme Court actually decide in March 2026?
It declined to hear Thaler v. Perlmutter, leaving in place lower-court rulings that human authorship is required for copyright. The practical effect is that an AI cannot be named as an author, and a work created entirely by AI with no human authorship is not eligible for protection. The decision did not ban AI-assisted creation.
03Why did Anthropic pay $1.5 billion?
To settle Bartz v. Anthropic, a case over the books used to train its models. It is the largest known copyright settlement in history. Companies are settling because the theoretical damages from losing a training-data case can run into the hundreds of billions — far larger than any settlement.
04Is AI-generated music safe to use commercially?
It is unsettled. Warner and Universal struck licensing deals, but Sony is still litigating the core fair-use question against Suno and Udio, with a pivotal ruling expected in summer 2026. Until then, the safest path is licensed AI-music platforms and a careful read of each tool's ownership terms before commercial release.
05Does this only apply in the United States?
No. The authorship rule is U.S. law, but the training-data fight is global. Germany's courts found OpenAI liable over lyric memorisation and will rule on GEMA v. Suno on July 31, 2026; the UK High Court has weighed in via Getty v. Stability AI; and trackers count 166 active cases across multiple jurisdictions. Outcomes differ by country, so check the law where you publish.
06What is the one thing I should do right now?
Start documenting human authorship on every AI-assisted project — your prompts, your edits, and the creative choices you made. It takes two minutes and is the single clearest evidence that a human, not a machine, authored the work, which is exactly what determines whether you own it.
