The global discourse surrounding artificial intelligence reached a critical juncture last week at the AI for Good Summit in Geneva, where industry leaders and content creators confronted a reality that has long been an open secret: the foundations of the current generative AI boom were built using vast quantities of proprietary data and creative content without explicit permission or compensation. Organized by the International Telecommunication Union (ITU)—the United Nations’ specialized agency for information and communication technologies—the summit served as a platform for a high-stakes debate over the future of intellectual property (IP) in an era where machines can replicate human creativity with increasing sophistication.
The admission that much of the AI industry has been constructed on "stolen" data has shifted the conversation from technical feasibility to ethical and economic sustainability. For major media conglomerates and legacy publishers, the challenge is no longer just about protecting their past work, but about ensuring a viable business model for the future.
Warner Bros. Discovery and the Protection of Creative Integrity
At the forefront of this defensive strategy is Warner Bros. Discovery (WBD), a media giant with a portfolio spanning from CNN and The Discovery Channel to DC Studios and Warner Bros. Entertainment. Avi Saxena, the CTO of Global Digital at WBD, articulated a firm stance during the summit regarding the company’s refusal to feed its vast library of content into public AI models.
According to Saxena, the company views itself primarily as a collective of creators. "Creators are at the center of the company, and we thrive on that," Saxena stated, emphasizing that the primary reason for withholding content from public models is to preserve the creative aspect of the industry. Instead of contributing to third-party platforms, WBD is developing its own internal models designed to be more ethical and specifically tailored to the needs of its creative community.
WBD has established rigorous internal guidelines to distinguish between different types of AI-assisted output. The company utilizes AI for what Saxena terms "derivative content"—the automation of promotional materials, trailers, and regional localizations. These tasks increase productivity and allow content to reach a global audience without infringing on the core creative process. However, Saxena drew a sharp line at "net new content," such as using AI to generate new episodes of established intellectual properties like Rick and Morty. By focusing on utility rather than creation, WBD aims to use AI as a tool for efficiency rather than a replacement for human talent.
The Transformation of Digital Publishing at TIME
While media conglomerates focus on protecting film and television assets, digital publishers are facing a more immediate crisis: the surge of automated traffic. Jessica Sibley, CEO of TIME, revealed a startling shift in the digital landscape, noting that her organization now observes more bot traffic on time.com than human visitors.
In response, TIME has been forced to engineer two distinct user experiences. The human experience remains focused on engagement, journalism, and advertising. Conversely, the bot experience is designed for restriction and monetization. Sibley noted that while TIME blocks a significant portion of bot traffic to protect its IP, it has allowed approximately 70 specific bots to access the site to participate in what she calls the "answer ecosystem" and Generative Engine Optimization (GEO).
Sibley’s concerns extend beyond simple data scraping; she highlighted a significant "delta" in how brands are represented by AI systems compared to their actual identities. Internal research suggests that the accuracy of how a brand’s values, products, and leadership are communicated via AI can vary by 15% to 40% depending on the source material the AI prioritizes. Sibley warned that if an AI system prioritizes a low-influence YouTuber over a 100-year-old trusted institution like TIME, the result is a "garbage in, garbage out" cycle that devalues high-quality journalism and damages brand reputations.
A Chronology of the AI-Content Conflict
The tensions displayed in Geneva are the result of a rapidly escalating timeline of legal and technological developments:
- 2020–2022: Large Language Models (LLMs) like GPT-3 are trained on massive datasets scraped from the open web, including news sites, books, and art repositories, under the assumption of "Fair Use."
- Late 2022: The public release of ChatGPT and Stable Diffusion brings generative AI into the mainstream, prompting immediate backlash from artists and writers.
- January 2023: Getty Images files a lawsuit against Stability AI, alleging the company "unlawfully" scraped millions of images to train its AI art generator.
- December 2023: The New York Times sues OpenAI and Microsoft, claiming that millions of its articles were used to train chatbots that now compete with the newspaper as a source of information.
- May 2024: Major licensing deals begin to emerge, such as the multi-year agreement between News Corp and OpenAI, valued at over $250 million, signaling a shift toward a "pay-to-play" model for premium data.
The Economic Power Imbalance
A central theme of the Geneva summit was the staggering economic disparity between the "Big Tech" firms developing AI and the fragmented landscape of content creators. Bill Gross, CEO of Prorata.ai, highlighted the difficulty of negotiations when five multi-trillion-dollar companies—including Microsoft, Alphabet (Google), and Meta—face off against thousands of individual publishers and studios.
Gross argued that while the collective value of all global brands and creators likely equals the $20 trillion market cap of the tech giants, their lack of a unified front prevents them from striking fair deals. "They’re fragmented, so there’s no negotiating power together at the table," Gross explained. He advocated for regulatory intervention to level the playing field, suggesting that the law must mandate compensation for the use of proprietary content.
Prorata.ai has proposed a revenue-sharing model where 50% of the income generated from AI-driven answers is distributed proportionally to the content owners. This model seeks to restore what Gross calls the "handshake deal" of the early internet. In the search engine era, Google crawled content but directed traffic back to the source. In the AI era, bots crawl content and provide the answer directly, effectively severing the traffic link and "breaking the deal" that allowed the digital economy to flourish for two decades.
Implications of Synthetic Data and Model Collapse
Beyond the legal and ethical arguments, there is a technical necessity for AI firms to reach agreements with content creators: the threat of "model collapse." As AI-generated content proliferates online, newer models are increasingly being trained on "synthetic data" (content created by previous AI versions) rather than human-generated data.
Experts at the summit suggested that a law of diminishing returns is beginning to take effect. If AI models continue to ingest their own output, the quality and accuracy of the systems will inevitably degrade. Avi Saxena of WBD noted that AI companies would benefit more from "another trillion tokens coming from proprietary content" than from public, low-quality data. This creates a "carrot" approach to negotiations: AI firms need high-quality, human-verified data to remain competitive, which may eventually force them to the bargaining table even in the absence of strict legislation.
The Path Toward a Sustainable AI Ecosystem
The consensus among the speakers in Geneva was that the current "wild west" era of data scraping is reaching its end. Whether through the "stick" of litigation or the "carrot" of licensing for better performance, the industry is moving toward a formalized royalty system.
Jessica Sibley of TIME expressed optimism that consumers would eventually seek out trusted sources over AI "slop," but emphasized that this trust must be supported by sustainable business models. Bill Gross echoed this sentiment, calling for a "morally right transaction" where tech giants like Apple or Google lead the way in establishing ongoing royalty payments rather than one-time legal settlements.
As the AI for Good Summit concluded, the message to the global tech community was clear: the next phase of AI development cannot be sustained on the unauthorized use of human intellectual property. The transition from a parasitic relationship to a symbiotic one will require a combination of technological innovation, such as Prorata’s attribution engines, and robust regulatory frameworks like the EU AI Act, which increasingly demands transparency regarding training data.
The resolution of these issues will likely define the media and technology landscape for the next century. For now, the battle lines are drawn between those who view the internet as a free resource for machine learning and those who believe that the value of human creativity must be protected and paid for in the digital age.
