Hundreds of Creatives Launch "Stealing Isn't Innovation" Against AI Training

Cate Blanchett, Scarlett Johansson, and R.E.M. join approximately 800 artists demanding licensing rights.

The website is stark: white text on black, no images, just names scrolling past like film credits. Cate Blanchett. Scarlett Johansson. Michael Stipe of R.E.M. By yesterday, around 800 creatives had signed onto "Stealing Isn't Innovation," a campaign that frames unlicensed AI training as theft at a grand scale, according to the Human Artistry Campaign's announcement.

The timing reveals either remarkable coincidence or strategic coordination. YouTube CEO Neal Mohan's acknowledgment of AI slop flooding the platform came just one day before the creatives launched their campaign. Yet Mohan simultaneously announced new AI tools for creators. This suggests YouTube recognizes the quality crisis but believes the solution is better AI tools, not fewer of them. In the same letter, Mohan announced new generative tools letting creators generate video backgrounds and soundtracks using AI, though PCMag reports the platform promises to label realistic altered footage.

"AI companies are exploiting practiced craft to amass billions," stated Dr. Moiya McTier, Senior Advisor to the Human Artistry Campaign, in materials shared with Mashable. She warns this creates an environment dominated by deepfakes and artificial avalanches of spam.

The campaign's demands sound simple, licensing and opt-out rights, but they challenge the entire foundation of modern AI development. Most major models were trained on scraped data without permission, making retroactive licensing nearly impossible. The existential stakes aren't hyperbole: if AI companies must pay for training data, development costs could increase exponentially, potentially reshuffling the entire industry hierarchy. According to Mashable's reporting, the group warns that profit-driven scraping threatens the entire U.S. entertainment industry and risks what they term AI model collapse. This death spiral occurs as AI-generated content floods the internet and future models increasingly train on synthetic rather than human-created data. Each generation becomes more generic and error-prone, eventually producing content so degraded it's unusable. Recent research suggests this isn't theoretical. It's already happening in some domains.

CalMatters tested major video generation models including Sora, Veo, and Kling against human dance choreography earlier this week. The dance choreography tests reveal artistic shortcomings, a critical weakness as models attempt to replace human output. While AI can mimic technical movements, it fails at culturally specific expression. These models excel at technical replication but fail at meaning-making. Cultural dances carry historical, spiritual, and social significance that can't be learned from visual patterns alone. This suggests AI may forever remain a sophisticated mimic rather than a true creative force.

The campaign frames this as more than a copyright dispute. The global soft power argument suggests that if AI models trained on American creative works flood global markets with synthetic content, the U.S. loses an authentic cultural voice. Homogenized AI output could replace the diverse, human-driven content that builds international influence and economic value.

Legislative efforts are pushing ahead: CalMatters notes that digital fingerprinting bills aim to protect performers' likenesses from unauthorized AI replication, though implementation details remain unclear.

YouTube's contradiction, acknowledging AI slop while accelerating AI features, reflects platform economics. The company likely calculates that creator-generated AI content, even if lower quality, costs less than licensing professional content or losing creators to competitors offering AI tools. YouTube appears to be betting that volume and convenience will outweigh quality concerns.

By avoiding technical solutions, the campaign makes a shrewd strategic choice. Technical fixes like watermarking or detection tools become obsolete as AI advances, but establishing the principle that unauthorized training equals theft creates lasting legal precedent. They're fighting for the rules of the game, not just this round.

The campaign demands opt-out rights for creators from AI training datasets, though enforcement mechanisms remain undefined. YouTube plans AI-generated backgrounds and soundtracks despite acknowledging content quality concerns. Legislative proposals include digital fingerprinting to protect performer likenesses.

The campaign's website remains active for additional signatures. The outcome likely depends on three factors: legal precedents from ongoing lawsuits, consumer tolerance for AI-generated content quality, and whether major platforms can develop AI tools that enhance rather than replace human work. If courts establish strong licensing requirements, or if audiences reject synthetic content en masse, the creators' resistance could prevail. Otherwise, economic pressure may force acceptance of an AI-dominated landscape.