ByteDance — the same company behind TikTok's ongoing data privacy saga — released Seedance 2.0 on February 10, and users immediately started churning out unauthorized deepfakes of Hollywood characters and A-list actors. Studios sent cease-and-desist letters. Unions condemned it. But the same Disney accusing ByteDance of IP theft signed a $1 billion deal with OpenAI two months earlier to put Mickey Mouse and Darth Vader into Sora. Meanwhile, buried under all the Hollywood noise: Seedance could clone someone's voice from just a photo of their face. That's the part your business should pay attention to.

On February 10, ByteDance — the company behind TikTok — launched Seedance 2.0. It's not the first AI video generator, but it's the most advanced one available right now.

Where previous models take a text prompt and give you a video clip, Seedance 2.0 accepts up to 12 reference files at once — images, video clips, audio — and produces 15-second clips with synchronized dialogue, ambient sound, and consistent characters throughout. It generates video and audio together in a single pass, which means lip-sync actually works. That's been one of the biggest giveaways in AI video up to this point, and Seedance mostly solves it.

Within days of launch, an Irish filmmaker named Ruairi Robinson typed a two-line prompt and got a fight scene between Tom Cruise and Brad Pitt on a destroyed rooftop. The clip hit over a million views on X. Other users made alternative endings to Stranger Things, Optimus Prime meeting Rocky in a fast food restaurant, and Kanye West singing in Mandarin inside a Chinese imperial palace. None of it was authorized. Most of it was convincing enough that casual viewers couldn't immediately tell it was fake.

The backlash from Hollywood came fast. On February 13, Disney sent a cease-and-desist accusing ByteDance of building "a pirated library of Disney's copyrighted characters." Paramount followed the next day, citing infringement across Star Trek, South Park, SpongeBob, and The Godfather. The MPA called it "unauthorized use of U.S. copyrighted works on a massive scale." SAG-AFTRA condemned it. The Human Artistry Campaign called it "an attack on every creator around the world." Warner Bros., Netflix, and Sony have since sent their own legal threats.

On February 16, ByteDance said it "respects intellectual property rights" and promised to strengthen safeguards. No specifics on what that means yet. The Seedance 2.0 API — originally scheduled to open globally on February 24 — has been delayed indefinitely while ByteDance works on content controls.

Deadpool co-writer Rhett Reese probably captured the mood best: "I hate to say it. It's likely over for us."

Before we go further, it's worth remembering who built this.

ByteDance isn't a random startup. It's the company that just spent two years at the center of the biggest data privacy fight in U.S. tech history. TikTok was briefly banned in January 2025 under the Protecting Americans from Foreign Adversary Controlled Applications Act. ByteDance spent $1.5 billion on Project Texas — an effort to route U.S. user data through Oracle — and former employees told Fortune that data continued going to Beijing anyway. The FTC sued ByteDance in August 2024 for flagrantly violating children's privacy law, alleging they knowingly allowed millions of kids under 13 on the platform while collecting their data.

TikTok's ownership finally changed hands in January 2026, with a new U.S.-based joint venture taking over. But even that deal has critics. The Atlantic Council published an analysis arguing that ByteDance retained control of TikTok's core algorithm through a licensing arrangement, and that the new structure "does little to alter the underlying risks." Meanwhile, TikTok's updated January 2026 privacy policy now collects precise geolocation data — a change from the ByteDance era, where the policy explicitly said it didn't.

Now this same company has built the most capable AI video generator in the world. One that, before ByteDance pulled the feature, could generate a realistic clone of someone's voice from a photograph.

I'm not saying don't use tools from Chinese companies. I'm saying know who you're dealing with, what data these tools collect, and what the track record looks like. That's just due diligence.

If you read our December piece on the Disney-OpenAI deal, the Seedance backlash shouldn't surprise you. Disney invested $1 billion in OpenAI and licensed 200+ characters to Sora — then sent cease-and-desists to everyone else doing the same thing without a deal in place. Google got lawyers. ByteDance is getting them now.

The playbook hasn't changed: partner on the IP holder's terms, or prepare for war. The MPA's cease-and-desist to ByteDance actually said the quiet part out loud — copyright infringement on Seedance is "a feature, not a bug."

If you're a business that owns any kind of intellectual property — brand assets, proprietary content, training data — the question isn't whether AI will touch your stuff. It's whether you'll have a deal in place when it does.

Our sponsor this week is Wispr Flow and let me tell you - I use it EVERY day. I can talk so much faster than I can type. Wispr Flow helps me get my thoughts out and gets me executing much faster than ever before. The best part - it learns my speech patterns and get’s better every time I use it.

Founders spend too much time drafting the same kinds of messages. Wispr Flow turns spoken thinking into final-draft writing so you can record investor updates, product briefs, and run-of-the-mill status notes by voice. Use saved snippets for recurring intros, insert calendar links by voice, and keep comms consistent across the team. It preserves your tone, fixes punctuation, and formats lists so you send confident messages fast. Works on Mac, Windows, and iPhone. Try Wispr Flow for founders.

Start flowing free

The copyright fight between Disney and ByteDance is interesting, but unless you own a movie studio, it doesn't land on your desk Monday morning.

This part might.

Before ByteDance pulled the feature on February 10, Seedance 2.0 could generate a realistic clone of someone's voice from a photograph of their face. Not a voice sample. A photo.

A Chinese tech blogger named Tim Pan demonstrated this live — uploaded a single headshot, and the model produced audio that matched his voice, speaking style, and intonation. No audio reference needed. ByteDance suspended the feature the same day after the backlash, calling it a risk to "the health and sustainability of the creative environment." Which is corporate-speak for acknowledging they shipped a feature that makes voice impersonation far too easy.

Every executive at your company has a headshot on LinkedIn. Many have photos on the company website, in press releases, at speaking events. If a model can infer vocal characteristics from facial structure — and Seedance showed it can — the barrier for deepfake fraud just dropped from "I need a voice recording" to "I need a profile picture."

This kind of fraud is already happening with cruder tools. Deepfake-enabled fraud hit $200 million in Q1 2025 alone. An engineering firm called Arup lost $25 million in a single incident after employees were fooled by deepfake video calls impersonating executives. These attacks work because they exploit something basic: you recognize the voice, so you trust the instruction.

ByteDance pulled the face-to-voice feature. But the underlying capability is out there now. Other models will build the same thing. Pulling one toggle on one platform doesn't undo the research.

So here's what I'd actually do about all of this:

Make deepfake verification a process, not a conversation. Your finance team and executive assistants need a rule: any "call from the CEO" requesting an urgent transfer gets verified through a second channel. Every time. No exceptions. And audit what executive content is publicly available — headshots, conference videos, podcast appearances. It's all source material now.

Be cautious with ByteDance tools. Seedance 2.0 is currently limited to Chinese users on ByteDance's Jianying platform, but a global rollout through CapCut — over a billion users — is expected soon. Before your team starts feeding content into it, ask: where does the data go? What gets retained? These are the same questions that triggered a federal law and a Supreme Court case around TikTok.

But don't sleep on the opportunity. The AI video market is projected to grow from $847 million to $3.35 billion by 2034 — driven by small and mid-sized businesses that can now produce professional video without a production team or a six-figure budget. Product demos, explainer videos, localized marketing, social clips. If you've been sitting out because video felt too expensive, the barrier just dropped.

One more thing — and this connects to our AI Bowl piece from two weeks ago. The Super Bowl showed that audiences react negatively when companies lead with AI in their marketing. Svedka's fully AI-generated ad was called "nightmare fuel." Google's Gemini ad — human-crafted creative, AI as the product — ranked #1. Same lesson applies here: use the tools, but lead with the human outcome. Keep the machinery in the background.

AI-generated video is getting exponentially better — and not just because of Seedance. Sora 2, Google Veo 3.1, Kling 3.0, Runway Gen-4.5 — the entire field has leaped forward in the past six months. Native audio generation, synchronized lip sync, 4K output, consistent characters across scenes. A year ago this stuff was a novelty. Now it's a production tool.

Seedance grabbed the headlines because of the Hollywood drama and the deepfake controversy, but the bigger story is the trajectory. These tools are all getting better, faster, and cheaper at the same time. That's the part that matters for your business.

Which means it's time to start building the workflows and processes to actually take advantage of it — not just marvel at the demos. Figure out where AI video fits into your content pipeline. Test the tools. Develop a point of view on what works for your brand.

But do it with guardrails. Know what data you're feeding into these platforms and where it goes. Know what your employees are experimenting with. And most importantly — know what your IP actually is. Not just the obvious stuff like logos and brand assets. Your training materials, your internal processes, your proprietary data, your executive likenesses. All of it has value in an AI-generated world, and if you haven't defined what's yours and what your stance is on protecting it, someone else will make that decision for you.

P.S: Please hit reply and let me know in one sentence if you found this week’s newsletter valuable.

Keep reading and learning and, LEAD the AI Revolution 💪

Hashi & The Context Window Team!