Alibaba's New AI Video Model Faces Tom Cruise Copyright Red Flag

Alibaba's latest AI video generation model, Wan3.0, just achieved general availability—and it's already embroiled in the same copyright controversy that has plagued the AI industry since its inception. The company's own promotional materials allegedly feature Tom Cruise's likeness without authorization, raising serious questions about whether the rush to deploy cutting-edge AI tools has outpaced the development of basic ethical safeguards and legal protections.

Alibaba's Wan3.0 represents a significant technical leap forward for AI-generated video. Unlike earlier iterations, the new model accepts multimodal inputs—text prompts, static images, video clips, audio files, PDFs, and even web pages—to generate coherent 30-second videos. The versatility is genuinely impressive from a capability standpoint. Users can now feed a model a simple image and a text description, and Wan3.0 will animate and expand that vision into a fully-formed video sequence. This flexibility opens legitimate creative possibilities for filmmakers, marketers, educators, and content creators operating within ethical and legal boundaries.

But technical capability and responsible deployment are two different things. When Alibaba showcased Wan3.0 to the world, the company's own marketing materials crossed a line that should have been obvious: they featured what appeared to be a deepfake of actor Tom Cruise, complete with his distinctive features and mannerisms, without any indication of authorization or consent. This isn't a subtle gray area. Using someone's likeness—celebrity or otherwise—without permission violates fundamental principles of personality rights and, in many jurisdictions, constitutes a violation of publicity rights and defamation law.

The Tom Cruise Problem Isn't New, But It's Still Happening

The irony here is thick. The AI industry has spent the better part of three years claiming to learn from previous deepfake scandals, and yet major corporations continue to repeat the exact same mistakes. The use of celebrity likenesses in unauthorized AI-generated content isn't novel—it's a recurring pattern. In 2023, deepfake videos of celebrities circulated widely across social media, and the industry responded with promises of better safeguards, watermarking, and authentication systems. Companies pledged they'd implement ethical guidelines and authentication mechanisms.

Alibaba's decision to feature what appears to be Tom Cruise in its promotional materials suggests those pledges may have been more marketing speak than genuine commitment. The Tom Cruise likeness—whether generated or manipulated—serves Alibaba's commercial interests perfectly: it demonstrates the model's capability to generate realistic human faces and movements. But that commercial benefit comes at the expense of one person's right to control their own image and likeness. If Tom Cruise didn't authorize this use, Alibaba's promotional strategy amounts to a public demonstration of unauthorized deepfake creation—which is precisely the kind of behavior that's made deepfakes a serious societal concern.

The broader implications extend beyond Tom Cruise or Alibaba specifically. This incident reveals a systemic problem within the AI industry: companies racing to market with new generative capabilities often treat copyright, personality rights, and ethical guidelines as afterthoughts rather than foundational requirements. The competitive pressure to launch first, launch big, and demonstrate maximum capability creates incentives to cut corners on legal and ethical review.

Training Data Opacity and the Foundational Problem

The Tom Cruise incident is symptomatic of a deeper issue that plagues the entire AI video generation space: training data opacity. Companies like Alibaba, OpenAI, Runway, and Pika rarely disclose exactly which datasets they used to train their models. Did Wan3.0 train on scraped internet video? Licensed content? Public domain material? Proprietary datasets? The vagueness is deliberate. Full transparency about training data would immediately expose the scale of copyrighted content, celebrity likenesses, and potentially non-consensual material that went into these models.

When you can't (or won't) tell the public what your model trained on, it becomes nearly impossible to audit whether the training process respected copyright, privacy, and personality rights. This lack of transparency creates a moral hazard: companies can claim they're committed to ethical AI while simultaneously building models on potentially unethical foundations. A model trained on thousands of hours of scraped celebrity video and images will naturally be very good at generating celebrity likenesses—which is exactly what makes it valuable commercially and exactly what makes it legally and ethically problematic.

The entertainment industry has taken notice. Industry groups and rights holders have increasingly questioned whether AI companies obtained proper licenses or permissions for the content used in training. The 2023 Hollywood strikes highlighted AI training data concerns prominently, with writers and actors demanding explicit protections against unauthorized use of their work and likenesses in AI training. Yet progress on concrete solutions has been glacial.

Alibaba's promotional use of what appears to be Tom Cruise suggests the company either obtained explicit authorization (unlikely, given the lack of any public statement from Cruise or his representatives) or proceeded without proper permissions. There's no third option that makes this scenario acceptable from a rights perspective.

The capabilities that make Wan3.0 impressive—its ability to generate convincing human figures, recreate facial expressions, and synthesize movement patterns—depend almost entirely on having trained on extensive real-world video and imagery. The better the model is at generating photorealistic people, the more likely it is that the model learned those capabilities from extensive training on celebrity footage, headshots, and performance footage. And those training dependencies create an uncomfortable relationship between capability and copyright infringement.

What separates responsible AI development from reckless development isn't just the capability—it's the safeguards built around that capability. Major AI labs have published responsible development frameworks, but Alibaba's apparent choice to feature an unauthorized celebrity likeness in its marketing material suggests that publication and actual implementation remain far apart. Adding guardrails after launch isn't good enough when the launch itself violates the principles those guardrails are supposed to protect.

The real test for Alibaba and other AI companies comes next. Will Wan3.0 include meaningful filters to prevent non-consensual deepfakes? Will the company implement authentication systems to track whether generated videos have been modified? Will Alibaba commit to transparent reporting about which training datasets were used and whether proper licenses were obtained? Or will the company follow the pattern established by many AI startups: launch with maximum capability, address ethical concerns only when forced to by regulatory pressure or legal action?

Until the AI video generation industry demonstrates a genuine commitment to addressing copyright, personality rights, and consent issues as core features rather than optional add-ons, incidents like the Tom Cruise problem will keep happening. Alibaba's Wan3.0 may be technically impressive, but it's also a reminder that in the race to build better AI, the industry is still running roughshod over the people—and the legal systems—that stand in the way.