Traditional Media Starts Talking Money with AI

A Long-Overdue 'Copyright Awakening'

When ChatGPT and its peers can fluently answer questions about news events, cite in-depth reports, and even recount exclusive investigations, a question has finally been laid on the table — where does all this content come from?

The answer is self-evident. Large language models scraped vast amounts of publicly available internet content during their training phase, a significant portion of which came from decades of high-quality news reporting, in-depth analysis, and opinion columns accumulated by traditional media. This content once underpinned the entire news industry's business model, yet it has now become the 'invisible fuel' driving AI companies' soaring valuations — all without authorization.

From 2024 to 2025, traditional media worldwide has finally begun collectively talking money with AI companies. This is not just a copyright defense campaign but a deep-seated contest between the traditional content industry and the emerging AI industry.

The Global Media 'Counterattack' Is Accelerating

This contest began with lawsuits. In late 2023, The New York Times took the lead by filing a copyright infringement lawsuit against OpenAI and Microsoft, alleging that they used vast amounts of news content to train AI models without permission. The lawsuit was seen as a landmark event signaling traditional media's declaration of war against the AI industry.

Similar legal actions have since spread globally. Multiple major Western media outlets have joined the litigation, spanning news agencies, newspaper groups, and broadcast organizations. Meanwhile, some media outlets have chosen a different path — sitting down directly with AI companies to negotiate partnerships.

Several major models have emerged within the industry:

• Licensing fee model: AI companies pay media organizations content licensing fees in exchange for the legal right to use news content for model training. The Associated Press, Axel Springer, and others have already reached such agreements with OpenAI.
• Revenue-sharing model: When AI products cite or generate answers based on media content, media organizations receive a proportional share of the revenue.
• Technology partnership model: Media and AI companies jointly develop news-oriented AI products, sharing technological outcomes and commercial benefits.
• Traffic referral model: AI platforms include links to original articles when citing news content, driving traffic to media websites.

Regardless of the model chosen, the core demand of traditional media has become crystal clear: you can use it, but you need a license; we can collaborate, but you can't take it for free.

Why Media's Bargaining Power Is Growing

Traditional media has been able to gradually seize the initiative in this contest thanks to several key factors.

First, the legal and regulatory environment is tilting in favor of content creators. The EU's AI Act and Digital Copyright Directive have provided a legal framework for media rights protection, explicitly requiring AI companies to obtain authorization or provide reasonable compensation when using copyrighted content. The U.S. Congress is also actively pushing related legislation, with multiple bills addressing AI training data copyright already under discussion.

Second, AI companies are beginning to recognize the importance of 'data compliance.' As industry competition intensifies and regulations tighten, legally obtaining high-quality training data has become a core competitive advantage for AI companies. Rather than facing hefty damages and reputational risks, it makes more sense to proactively seek licensing partnerships. Tech giants including OpenAI, Google, and Apple have signed content licensing agreements with multiple media organizations in recent years, which itself speaks to the shifting trend.

Third, the scarcity of high-quality content is becoming increasingly apparent. As AI-generated content floods the internet, news reports written by professional journalists and subjected to rigorous editorial processes are becoming even more scarce and valuable. If AI models are trained solely on AI-generated content, they risk falling into a 'model collapse' trap, further elevating the value of authentic, high-quality human-created content.

Challenges and Controversies Coexist

However, this journey of talking money has not been entirely smooth sailing.

The pricing dilemma is the biggest obstacle. How much is media content actually worth to AI training? There is currently no unified valuation standard in the industry. Disclosed licensing agreement amounts vary enormously, ranging from several million to tens of millions of dollars per year, with vastly different pricing rationales. Large media groups hold stronger bargaining power, while small and mid-sized outlets risk being marginalized.

The boundaries of 'fair use' remain blurred. AI companies broadly argue that using publicly available data to train models falls under 'fair use' and does not require additional authorization. This position has yet to be definitively settled in law, and court rulings across different countries may reach entirely different conclusions, introducing uncertainty into industry-wide rule-making.

The fairness of benefit distribution also deserves attention. In licensing agreements reached so far, the beneficiaries tend to be top-tier media institutions. How to protect the rights of numerous local media outlets, independent journalists, and freelance writers remains a question without effective mechanisms in place.

Additionally, some media professionals worry that once AI companies secure sufficient licensed content, they may further accelerate the replacement of traditional media. Could partnership agreements turn out to be a 'sweet poison'? These concerns are far from unfounded.

Exploration and Practice in Chinese Media

Domestically, the copyright contest between traditional media and AI companies is also heating up. Multiple mainstream media outlets and content platforms have begun discussing content licensing partnerships with AI companies such as Baidu, Alibaba, and ByteDance.

Since 2024, several domestic media groups have signed strategic cooperation agreements with large model companies, covering areas such as news content licensing and joint product development. Institutions like the Copyright Protection Center of China are also actively working to establish copyright registration and transaction mechanisms for AI training data.

Notably, regulatory bodies including the National Copyright Administration of China have repeatedly emphasized that using others' works for AI training should be done with proper authorization under the law, providing policy support for media rights protection.

Outlook: From Confrontation to Co-Building

Looking from the vantage point of 2025, the relationship between traditional media and AI companies is gradually shifting from initial 'confrontation' toward 'conditional cooperation.'

In the short term, more licensing agreements will be reached, and industry pricing standards will gradually take shape. In the medium term, dedicated content licensing trading platforms and standardized copyright management tools may emerge, making the licensing process more efficient and transparent. In the long term, traditional media and AI companies may establish a new mutually beneficial symbiotic relationship — media provides high-quality content, AI provides technology and distribution capabilities, and both sides jointly create new commercial value.

But the prerequisite is that rules must be clear, rights must be respected, and value must be acknowledged.

As one veteran media professional put it: 'We don't oppose technological progress, but technological progress cannot be built on the unpaid exploitation of content creators.'

This contest over AI and content copyright has only just begun in earnest.