The world has changed a long time ago.

In the past few years, from seeing my colleagues using AI to quickly produce an excellent technology media article to having colleagues even using AI to imitate my word choice, rhythm, and writing logic, I have basically been unable to suggest any revisions. AI can leapfrog humans almost instantly.

At the same time, today I am creating videos like a director every day, trying different film and television and aesthetic styles; using the ability of Agent to write software, expand the world view, and gain professional judgment that requires years of accumulation.

In the past few years, "craftsmanship" has been losing and growing at the same time. This high-speed energy convection is truly shocking as never before.

There is no doubt that those abilities that have taken years of hard work to internalize seem to be "algorithmic" in an instant in front of AI, and all those "unique" seem to be disappearing. Our historical accumulation is becoming "nothing", but because of this, everyone seems to be able to do "anything" in the future.

This incident extended to the commercial system of human society, and a direct impact was the beginning of the collapse of "copyright". This is a problem that seems to be urgently needed to be solved. But in fact, it is likely to have entered an irreversible process of disintegration that is no longer a problem.

In fact, the disputes surrounding "copyright" have also extended from the issues of AI training and data use to whether content generated by AI enjoys copyright? If AI is used to generate content that is similar to the copyright owner’s work, does it constitute infringement? An "uncertain state" in which problems continue to expand but remain unresolved.

In fact, if we extend the timeline, we will see that copyright issues may no longer be viewed and solved from a historical perspective. We will see a more continuous process: the foundation of the copyright system did not just start to shake today, but has been weakened and reconstructed layer by layer over the past twenty years or so until it entered a completely different state. The disintegration of copyright is a process that has already begun and may be irreversible. It is a structural collapse of the concept of copyright in a general sense.

Back to the source: Why did the copyright system come into being?

In order to better understand this collapse, we have to go back to the starting point: How was the copyright system born?

As early as the Song Dynasty, when engraving and printing replaced handwriting and greatly increased the speed of copying and disseminating books, publishers became aware of copyright and formed the concept of "copyright". Copyright in English is copyright, which literally means "right of reproduction". It was also born in the context of the emerging printing and publishing industries. For publishers, the most important thing is to control the printing and publishing rights of their works.

The core protection of copyright is the originality of the work. What it defends is not the quality of the work, but its "uniqueness". By granting creators exclusive control over the act of copying for a specific period of time, it creates an artificial scarcity in circulation, thereby ensuring that high intellectual investments receive the commercial returns they deserve. Copyright itself is also a product of a technological revolution.

The foundation of the copyright system is established in a physical world where physical copying is relatively controllable and has a certain cost, and information dissemination is relatively slow. However, rounds of technological innovations continue to impact the copyright wall.

1.0 Search Era, "Indexable": The First Transfer of Copyright

In the early days of the Internet, the copyright wall ushered in the first wave of erosion. In order to provide effective retrieval services to users, search engines represented by Google use their technical mechanism to crawl, cache, and index all public web pages on the entire network. Because every step involves copying original content, this series of operations triggered a lot of copyright disputes around 2000: Does unauthorized copying and indexing of content constitute infringement?

In 1994, the "Crawler Protocol" (robots.txt) was proposed by Martijn Koster and formed an industry self-regulatory standard. Websites can declare the scope of crawling, and mainstream search engines voluntarily comply, forming an industry practice of "tacit approval but rejection". In 1998, the United States passed the Digital Millennium Copyright Act (DMCA), in which the "Safe Harbor Principle" stipulates that as long as the platform acts as a "passive intermediary" and responds to deletion notices in a timely manner, it can be exempted from infringement liability. This leaves room for "infrastructure roles" such as search engines and hosting platforms.

A landmark case in judicial practice on copyright issues was the "Kelly v. Arriba Soft Corp." case that occurred in the United States at the turn of the century. Photographer Leslie Kelly believed that Arriba's image search engine infringed her copyright by appearing in search results without permission. In that case, the court ultimately ruled that thumbnails produced by search engines to generate previews were fair use because they were transformative—the function was transformed from artistic expression to information indexing, providing the public with new value for easy retrieval without harming the market for the original work. This ruling established an important legal precedent for the search engine industry's operating model.

The essence of this stage is that the content party gives up part of the control rights of the "copied" for the opportunity of being "discovered". In the ocean of information, the anxiety of "the smell of wine is also afraid of the depth of the alley", overwhelms the persistence of the right to copy. After all, the ultimate purpose of copyright is to achieve value, and uniqueness that no one cares about is worthless.

Represented by the "Crawler Protocol" and the "Safe Harbor Principle", a series of rules have emerged. They do not deny copyright, but use technical patches to transform the original solid walls into conditional access gates. For the first time, the boundaries of copyright were penetrated by the logic of technology. It still exists, but it has begun to serve a larger goal: visibility.

2.0 Information Flow Era, "No More Jumping": The Second Transfer of Copyright

If the search engine era still respects "links" as a symbol of content ownership, then the rise of information flow platforms has completely changed the way content exists and is consumed. They are no longer just "entrances" to original content, but have directly become "containers" for containing content. The entire process of users consuming information, from discovery, reading to interaction, is completed within the platform, without the need to jump to the original website.

In the early days of Toutiao’s founding, it used technical means to automatically transfer content from other online media to Toutiao’s platform, which aroused widespread dissatisfaction and collective rights protection in the media industry. Despite the turmoil, traditional media has declined irreversibly, and content platforms have become the norm.

This "containerized" content consumption model triggered fierce industry conflicts in the early stages of the development of information flow platforms. The early development history of Toutiao is the most typical example. In the early days of its establishment, Toutiao did not invest a lot of resources in the production of original content. Instead, it relied on its core algorithm to batch crawl and aggregate news, information and other content from other online media, automatically transfer it to its own platform and distribute it, so as to quickly accumulate users and seize the market. This model directly bypasses the copyright ownership of the content and breaks the industry consensus of "links point to the original" in the search engine era. Therefore, it quickly aroused dissatisfaction among the media industry and jointly initiated collective rights protection. Although this copyright controversy ultimately ended with Toutiao adjusting its content strategy and gradually introducing legitimate cooperation, it also clearly reflected the inevitable change of the industry - the platform model of "content aggregation + algorithm distribution" has become the industry norm in the era of information flow, paving the way for subsequent structural changes in the content ecosystem.

This change has brought about two profound structural impacts. First, distribution rights have become a core resource in the content ecosystem. When users no longer visit the original site by clicking on the link, the visibility of the content no longer depends on the influence of the original website itself, but entirely depends on the recommendation of the platform algorithm. The traffic gate is firmly controlled by the platform. Media organizations that once collectively denounced Toutiao eventually chose to join in large numbers, and even took the initiative to operate accounts on multiple platforms to submit works, transforming from copyright owners into "content factories" that rely on channel distribution.

Secondly, the more far-reaching impact is that it destroys the original business model of content creators-hard advertising. In the era of search, traffic will eventually be directed to the creator’s own website, and the advertising space on the website is an asset that can be sold independently. But on the information flow platform, all advertising spaces on the page belong to the platform, and creators cannot sell them independently. Its commercial monetization path has been compressed into the platform sharing and accepting "soft article advertisements", and whether soft articles can be recommended by algorithms is still full of uncertainty. Creators not only need to invest in producing high-quality content, but in order to get the content seen by more people, they even need to pay extra to purchase traffic through the platform's commercial promotion tools. The identity of the creator has been downgraded from a subject with an independent business closed loop to a production unit in the platform ecosystem, and the commercial value of his original business attention has been concentratedly harvested by the platform. The economic significance of copyright was further eclipsed at this stage.

3.0 model era, "internalized as ability": the third transfer of copyright

Generative AI, represented by ChatGPT, Midjourney, etc., has once again brought about fundamental changes in the way content is processed.

In the search era, content is "link"; in the information flow era, content is "distribution"; and in the model era, content becomes "weight" in probability distribution. The way the big model handles content is no longer "linking" or "distributing", but "digesting, absorbing and integrating". It reads massive amounts of text, images, code and other data, decomposes it into the most basic knowledge, patterns and parameters, and internalizes it as part of its own neural network. Then, when users ask for it, it can generate content in a new, seemingly original way.

Under this model, the copyright value of any independent report, a well-taken photo or a piece of code has been diluted as never before. They are no longer cited as a complete "work", but as one of the countless raw materials that make up the model's intelligence. Your content may have had a one-in-a-trillion impact on a certain parameter of the model, but you can’t trace it back, and it’s even harder to claim compensation.

If copyright is to protect "originality," then when AI can quickly learn and imitate any style and any brushwork, "originality" itself becomes fragile. Not only have creators lost their business model, but their production value as a "content factory" has been drained from the bottom of the cauldron. The foundations of copyright have been completely loosened.

It can be deduced that what may happen in the future is that in order for their content to be more likely to be presented in AI-generated answers after being captured and absorbed by large models, and thereby indirectly gain exposure, traffic and even commercial value, creators have to invest additional energy and cost - the business of GEO (Generative Engine Optimization) has also been pushed to the forefront. Learn and apply GEO related techniques, optimize the structure, keywords and presentation form of your own content, and adapt to the capture and learning logic of large models. Just like creators pay to purchase traffic in the information flow era, GEO has become the era of generative AI, where creators compete for a new track to be "seen by AI", and a cycle of "passively adapting to new rules and paying for exposure" has officially begun, further reshaping the underlying logic of the content ecosystem.

An asymmetric reality: "Copyright Aristocrats" and "Copyright Commoners"

The loosening of the copyright wall does not mean that a completely equal copyright-free era will immediately usher in. At least for a certain period of time, it intensified the division of the world of creators.

A few "strong copyright entities", such as Disney and large news groups, who own massive, high-quality, structured data and have strong legal and brand strength can still compete with AI companies at the negotiation table as "copyright aristocrats". The bargaining chip they hold makes their copyrights still an asset that can be priced. The New York Times, for example, sued OpenAI and Microsoft, accusing them of using millions of its articles without permission to train AI models. Similarly, giants such as Disney and Universal Music Group also have the ability to sit at the negotiation table with AI companies and seek business cooperation or licensing agreements with their large content libraries and legal teams. Their struggles are the focus of the media, but their paths are difficult for most ordinary creators to replicate.

But for the vast majority of individual bloggers, self-media, artists, programmers, and small and medium-sized content agencies, they do not have the bargaining chips to compete with AI giants, and the protection of copyright is disintegrating.

Looking back at these three stages, we see a clear trajectory: the boundary of copyright continues to recede, and the core of power has changed from "controlling copying", to "link discovery", to "monopoly distribution", and now to the "training contribution" that is difficult to measure today. The deeper logic behind this is that every iteration of technology is a redistribution of attention control.

In the content industry, human beings’ limited attention is the only source of value. The original intention of the copyright system is to indirectly protect creators' attention and profits by protecting the "originality" and "right of reproduction" of works in an era of scarce information. However, the evolution of technology is essentially a process of "unbinding" attention from original works and "scheduling" and "aggregating" them on a large scale. Search engines allocate attention to the discovery process, and information flow platforms focus attention on the consumption process. The large AI model attempts to directly dissolve the source of attention - it commercializes the "smart model" itself that attracts attention, deconstructs the value of the smart model behind content creation, and abstracts it into a capability that can be infinitely personalized and integrated into the production environment. When the foundation that supports the value of copyright—that is, the stable control of attention—is taken away again and again, the upper building will naturally collapse.

For creators, sticking to the "copyright" of historical works is probably not the answer. How to find your own value coordinates in the new smart ecosystem will be a topic that every creator needs to think about.

At the same time, in the face of this reality, we must realize that the rules of the high-speed digital world are being rewritten. With the information processing capabilities and technological iteration speed of AI, it will become increasingly difficult to build barriers in the digital world. Not only content copyright, but also patent barriers in the industrial field may also face the same challenges. AI can quickly analyze existing patents and design numerous evasion routes. In the future, truly effective barriers may only exist in the low-speed physical world—areas that cannot be easily digitized and copied, such as complex supply chain management and control that takes time to settle, interpersonal trust based on in-depth interaction, and real offline experiences that cannot be replaced.