Platform news and rules

Hollywood and ByteDance just signed an AI copyright deal. Here's what it covers.

On August 18, 2026, the Motion Picture Association and ByteDance announced a framework agreement to add copyright guardrails to CapCut, Dreamina, and TikTok's AI image and video models. The deal caps roughly six months of tension that began when a fabricated fight scene between Tom Cruise and Brad Pitt, built with ByteDance's AI tools, spread across TikTok and prompted a cease-and-desist letter from the MPA. It's a narrow agreement between two specific parties, not a new law, but it's a useful signal for anyone using AI video tools to build content.

What the agreement actually covers

The framework applies to ByteDance's consumer-facing tools, CapCut and Dreamina, along with TikTok's own image models, with the MPA noting that newer versions of ByteDance's Seedream (image) and Seedance (video) models already demonstrate stronger protections. MPA Chairman Charles Rivkin described it as putting "meaningful guardrails" in place to protect studio work; ByteDance's General Counsel John Rogovin called it "an important framework for continued collaboration" as the tools keep evolving. Reported safeguards include blocking generation from inputs containing real faces and blocking unauthorized generation of clearly copyrighted characters, alongside visible watermarks and embedded C2PA content credentials on outputs, the same provenance standard behind YouTube's own automatic AI-content labeling.

A framework, not a law

This is a memorandum of understanding between one AI toolmaker and one trade association, not legislation, and it doesn't create new legal obligations for a creator using these tools. The underlying fair-use questions the deal is meant to defuse are still being litigated elsewhere; studios continue pursuing separate suits against other AI companies, Midjourney included, over similar ground. What the deal signals matters more than what it legally requires: the direction the whole AI-video industry is heading is toward tighter controls on generating recognizable copyrighted characters and real people's likenesses without consent, echoing the same shift covered in the EU AI Act's labeling rules.

Why this matters even if you don't use ByteDance's tools

The consent and likeness principles at the center of this deal are the same ones our legal and policy basics guide already recommends treating as a hard line: don't generate a real person's face or a studio's copyrighted character without a license, regardless of which tool makes it technically possible. A major toolmaker agreeing to build that restriction directly into its products is a strong signal that platforms and courts are converging on the same standard, whichever tool a channel happens to use.

Where Thothium fits

Thothium generates original, scene-specific visuals from your own script, never recreations of a studio's copyrighted characters or a real person's likeness, so the guardrails this deal is pushing the industry toward are already the default rather than an added restriction. It is in free alpha, and the form below gets you a key.

Frequently asked questions

Does this deal change what I can do on YouTube?

Not directly. The agreement is between the Motion Picture Association and ByteDance, covering ByteDance's own tools and platforms, CapCut, Dreamina, TikTok's image and video models, not YouTube or any Google product. It's still worth knowing, because it shows where the industry as a whole is heading on AI-generated likeness and copyrighted-character content.

What exactly triggered it?

A viral deepfake video depicting a fabricated fight scene between Tom Cruise and Brad Pitt, made using ByteDance AI tools and spreading widely on TikTok, prompted the MPA to send ByteDance a cease-and-desist letter earlier in 2026. Months of negotiation followed, ending in this framework agreement.

Does this mean CapCut and Dreamina now block generating real people or copyrighted characters?

The agreement calls for guardrails against generating recognizable copyrighted characters and against real-face-based deepfakes, with newer versions of the Seedream and Seedance models already showing stronger protections. The exact technical implementation across every ByteDance product is still ByteDance's to roll out, and the public announcement did not spell out every detail.

Is this legally binding?

It's a framework agreement, a memorandum of understanding, not legislation, and it doesn't create new legal obligations for creators using these tools. What makes it notable is that it's the first agreement of its kind between a major AI-video toolmaker and Hollywood's trade association, and studios are still separately suing other AI companies, Midjourney among them, over similar fair-use questions the deal doesn't resolve industry-wide.

Last updated August 31, 2026. This is general information about a corporate agreement, not legal advice; the deal does not create new legal rights or obligations for creators, and specific implementation details across ByteDance's products may change as the framework rolls out.

Original scenes, not borrowed likenesses

Thothium generates original per-scene visuals from your own script, not recreations of copyrighted characters or real people's faces. Free alpha.
free alpha · no credit card · no spam