AI companies didn't just use creators' content. They allegedly broke through YouTube's security systems to get it. And right now, 97 active copyright lawsuits are testing whether that distinction changes everything.
Most conversations about AI and copyright come down to one question: is it legal to train an AI on someone's content? That debate has produced mixed results in court. But a new cluster of lawsuits is asking a different question, one that could be more powerful: does it matter how the data was collected?
The answer could be more consequential for creators than anything that has come before it.
Who Is Ted Entertainment, and Why Does It Matter?
Ted Entertainment is the company behind h3h3Productions, one of YouTube's biggest creator channels. Together with Matt Fisher (MrShortGameGolf) and Golfholics Inc., they've filed a coordinated series of lawsuits against the largest AI companies in the world. The claim isn't just copyright infringement. It targets a specific legal provision most creators have never heard of: DMCA Section 1201.
This isn't a single case. It's a coordinated campaign designed to test one specific legal theory across multiple defendants at the same time.
| Date | Development |
|---|---|
| December 2025 | First suit filed against Nvidia, alleging scraping of YouTube content to train the Cosmos video model |
| January 2026 | Suits extended to Meta, ByteDance, and Snap, all alleging the same methods: IP rotation, virtual machines, and bypassing YouTube's protection systems |
| April 2026 | New complaints filed against OpenAI, Apple, and Amazon. Seven total lawsuits now in progress, with more expected |
Image source: Shutterstock.
What Is DMCA Section 1201?
Most creators know the DMCA as the takedown tool. That's Section 512. You use it when someone uploads your content without permission. Section 1201 is different. It's the anti-circumvention provision.
In plain language: Section 1201 makes it illegal to bypass any digital lock that protects copyrighted content. You don't need to prove someone copied your content. You only need to prove they broke through the lock to access it.
That distinction is everything.
| Standard Copyright Infringement | DMCA §1201 Anti-Circumvention | |
|---|---|---|
| Based on | Unauthorized copying of content | Bypassing access controls |
| Fair use defense | Available | Does not apply |
| Violation occurs | At point of output | At point of access |
| AI track record | Companies have partly succeeded here | Largely untested until now |
Section 1201 carries its own penalties: up to $2,500 per act of circumvention in civil cases, with criminal penalties available for willful violations.
The YouTube "Rolling Cipher": What Was Actually Broken
Think of YouTube like a vault with a combination lock that changes every few seconds. The outside is visible to everyone: the thumbnail, the title, the view count. But the actual video file, the audio track, the raw frames? Those sit behind an encryption system called the "rolling cipher," built specifically to stop anyone from bulk downloading video at scale.
There is no legitimate way to mass-download YouTube video files. That's not an oversight. It's a deliberate barrier.
According to the complaints, AI companies didn't just scrape what was publicly visible. They used automated programs and virtual machines to continuously rotate IP addresses, evade YouTube's detection systems, and extract raw video and audio data to train their models.
They didn't just take the content. They broke the lock first.
This argument has already passed its first legal test. In a related case against AI music generator Udio, a federal court in the Southern District of New York found that the rolling cipher circumvention allegations were strong enough to proceed to trial.
Why This Strategy Is Different From Everything Before It
AI companies have been good at defending copyright claims using fair use. They argue that training on existing content is transformative, and some courts have agreed.
Section 1201 cuts around that defense entirely. The violation happens at the moment of access, not the moment of output. That means AI companies can't argue their way out with "but we transformed it." The question isn't what they did with the content. It's how they got it.
With 97 active copyright lawsuits against AI companies on the docket in the US right now, the outcomes of these cases will shape creator rights for the next decade.
Three Outcomes That Could Reshape Everything
1. Machine unlearning becomes a legal remedy
Courts could order AI companies to remove the influence of unlawfully obtained content from their models. This process, called machine unlearning, is technically complex and extremely expensive. That's exactly what makes it a powerful deterrent, and why AI companies will fight hard to stop it becoming standard practice.
2. Fair use boundaries are redrawn
A series of wins under §1201 would establish that how you collect training data matters as much as what you do with it. Transformative use wouldn't be enough. AI companies would need to prove they obtained data legally.
3. Licensing becomes the default
Since removing data from trained models isn't practical at scale, the most likely industry response is a move toward formal licensing agreements. This is already happening with major platforms and media companies. For individual creators, it points toward a future where your content has a documented ownership trail that AI companies must license rather than bypass.
What Creators Should Do Right Now
These cases won't resolve fast. Federal copyright litigation takes years, and every major ruling will be appealed. But you don't need to wait for the courts to act.
The creators who will benefit most from favorable rulings are the ones who already have a documented, verifiable record of what they created, when, and where it has appeared without permission.
- Document your ownership. Timestamped records, platform publication history, and copyright registration where possible. This is the foundation of any future legal claim.
- Set up content monitoring. Unauthorized use needs to be caught and removed quickly. Every takedown you file adds to your documented record.
- Know your platform's protections. If your content sits behind a paywall, a login, or any access control system, it may already qualify for §1201 protection.
If you create content behind a paywall or access-controlled platform, the protections being tested in these lawsuits may apply directly to your situation. Document your content ownership now.
Key Terms You Should Know
| Term | What it means |
|---|---|
| DMCA Section 512 | The standard takedown system, used when someone posts your content without permission |
| DMCA Section 1201 | Makes it illegal to bypass digital locks on copyrighted content, regardless of what you do with it afterward |
| Technological Protection Measure (TPM) | Any digital system that controls access to content: paywalls, encryption, login gates, rolling ciphers |
| Rolling cipher | YouTube's rotating encryption that prevents bulk downloading of video files |
| Fair use | A legal defense for limited use of copyrighted material without permission. Does not apply to §1201 claims |
| Machine unlearning | Removing training data's influence from an AI model. Can be ordered as a court remedy |
| IP rotation | Cycling through different internet addresses to avoid detection while scraping content at scale |
| Copyright registration | Official recording of ownership with the U.S. Copyright Office. Required to claim statutory damages in court |
| Statutory damages | Fixed damages set by law. $2,500 per §1201 violation, no need to prove specific financial harm |
| Training data | The content AI companies use to build their models, and the center of most AI copyright litigation right now |
Frequently Asked Questions
Is it legal for AI companies to use YouTube videos for training?
This is currently a contested legal question. AI companies often claim fair use for training on publicly available content. But the Ted Entertainment lawsuits argue that bypassing YouTube's security systems to access that content violates DMCA Section 1201, making the extraction illegal regardless of how the AI uses it. Fair use does not apply as a defense to Section 1201 claims.
What is the difference between copyright infringement and DMCA Section 1201?
Copyright infringement is about unauthorized copying. AI companies often defend against it using fair use arguments. Section 1201 is about unauthorized access. It makes it illegal to bypass digital locks to reach copyrighted content, and fair use is not a valid defense. The violation happens at the point of access, not the point of output.
Can machine unlearning remove my content from an AI model?
In theory, yes. A court could order an AI company to erase the influence of unlawfully obtained training data from its model. In practice, this is technically very difficult and expensive, which is why AI companies are fighting hard to prevent it from becoming a standard remedy.
How can creators protect their content from unauthorized AI use?
Start with documentation. Build a clear, timestamped record of your content ownership and publication history. Use a monitoring service like Rulta to detect unauthorized use automatically, enforce takedowns, and maintain the verifiable paper trail you'll need for any future legal claim or licensing dispute.
What This Means for Rulta, and for You
Rulta's job is removing your content from places it shouldn't be. What these lawsuits are making clear is that content protection and content documentation are now the same problem.
The infrastructure we operate today, continuous monitoring, automated detection, and verifiable removal records, is built on the same legal foundation being constructed in federal courts right now. When §1201 claims mature, courts will ask exactly the questions our system is already built to answer: when was this content published, by whom, and where has it appeared without permission?
That paper trail isn't just useful for takedowns. It's the foundation of any future legal claim, whether that's a standard infringement case, a §1201 claim, or an AI training data licensing dispute.
Owning your content, documenting it thoroughly, and acting quickly when it appears without permission are not just best practices anymore. They are the preconditions for whatever rights courts ultimately recognize in the AI era.
We'll keep tracking the Ted Entertainment cases and the broader AI copyright landscape. As things develop, we'll tell you what it means for you and how your protection strategy should evolve.
Start Protecting Your Content
Join the creators who are already protecting their work.
This post is for informational purposes only and does not constitute legal advice.
Sources: Ted Entertainment Inc. v. Apple Inc., Case 3:26-cv-02936 (N.D. Cal., filed April 3, 2026); Ted Entertainment Inc. v. Amazon.com Inc., Case 2:26-cv-01134 (W.D. Wash., filed April 3, 2026); Tyz Law Group, "Beyond Infringement: The Rise of DMCA Claims in AI Litigation" (March 2026); Sher Tremonte, "For AI Companies, How You Get Your Training Data Matters" (May 2026); chatgptiseatingtheworld.com Master Chart of AI Copyright Claims (April 4, 2026); Norton Rose Fulbright, "AI in Litigation: Update on AI Copyright Cases in 2026" (March 2026).
