Six weeks ago, a video featuring Tom Cruise and Brad Pitt fighting on a rooftop went viral. The footage wasn’t real but was created by Seedance 2.0, an AI video model from ByteDance. This incident led to cease-and-desist letters from six major Hollywood studios, a formal denunciation from the Motion Picture Association, and a rebuke from SAG-AFTRA regarding unauthorized use of member likenesses. Rhett Reese, the Deadpool screenwriter, commented on the tech’s implications for his profession.
ByteDance is now attempting to relaunch the tool with enough safeguards to address the backlash. On Wednesday, the TikTok parent company announced that its global safety and intellectual property teams, alongside a third-party red-teaming partner, enhanced Seedance 2.0 for its international release via CapCut, ByteDance’s video editing platform with over 400 million monthly users.
The new safeguards are significant. Seedance 2.0 now prevents video generation from real faces, tackling the deepfake controversy of February. CapCut will also block unauthorized generation of copyrighted characters, addressing previous AI-generated characters that the MPA cited.
For transparency, all outputs will feature visible watermarks and embedded C2PA Content Credentials for identifying AI-generated content. ByteDance introduces “advanced invisible watermarking” to identify model-generated content even after sharing or off-platform alteration and promises proactive monitoring for IP violations.
The rollout displays calculated caution. Seedance 2.0 will initially be available to paid users in Brazil, Indonesia, Malaysia, Mexico, the Philippines, Thailand, and Vietnam. Notably absent are the US and India. Europe, Africa, South America, and Southeast Asia will follow, though no firm US timeline is provided.
The AI video arms race is heating up. Just days ago, OpenAI announced the shutdown of its AI video tool, Sora, after download drops and a failed Disney deal. ByteDance, however, is moving forward into a market now acutely aware of AI-generated content’s regulatory issues.
The EU AI Act’s 2026 transparency rules will require generative AI providers to mark outputs in machine-readable formats and disclose deepfake origins. ByteDance’s watermarking may anticipate these requirements, but acceptance from European regulators is uncertain.
Red-teaming reports indicate the guardrails aren’t foolproof. Creative prompts can bypass filters, creating “likeness-adjacent” content that suggests real or copyrighted figures without exact duplication. This challenge persists in AI governance: the gap between policy and model output.
ByteDance’s vertical integration provides a unique position. It owns the AI model, the deployment platform CapCut, and TikTok, the leading short-form video platform. This control theoretically allows IP protection enforcement across the entire pipeline. Whether ByteDance will sufficiently satisfy Hollywood and its lawyers remains to be seen.
The AI boom of 2025 produced tools generating text, images, and code. Video stands as the next frontier, challenging to govern. ByteDance bets it can commercialize AI video generation globally without extensive litigation. The added safeguards to Seedance 2.0 are necessary initial steps. Whether they suffice is a question Hollywood, regulators, and policymakers will explore in the coming months.