The escalating dispute between the US news industry and Open

In the United States, the generative AI technology represented by ChatGPT is changing rapidly. However, AI cannot be completely "invented out of nothing". Massive Internet information is the fertile ground for its cultivation. This can easily lead to disputes over intellectual property rights. The lawsuit between media plaintiffs represented by The New York Times and OpenAI, the Open Artificial Intelligence Research Center, has been referred to by the media as a case of "defining AI copyright".

 

Starting from the end of 2022, OpenAI's artificial intelligence product ChatGPT will become popular. Internal testing by The New York Times found that ChatGPT can replicate news in large sections of the original text, directly replacing paid subscription reading and providing it to users; The New York Times has contacted OpenAI multiple times, requesting to stop crawling and sign a content authorization agreement, but negotiations between the two sides have broken down.

 

At the end of 2023, The New York Times officially sued OpenAI, its investors, and partner Microsoft in court.

 

New York Times Company President and CEO, Kopit Levin: We believe that they should pay a fair value exchange fee for the content they use to build their models. These artificial intelligence companies have spent hundreds of billions of dollars on computing power, energy, and talent. This is not only a problem faced by news organizations, but also by anyone engaged in creative work and any intellectual property creator.

 

However, OpenAI responded that the use of public Internet materials to train AI models was "reasonable use" and existed for a long time. This principle is fair to creators and necessary for innovators.

 

Subsequently, multiple media outlets filed lawsuits against OpenAI, and the Manhattan Federal Court merged multiple lawsuits into one.

 

Recently, the focus of the lawsuit is that the plaintiff is requesting OpenAI to use a search function to provide chat and training data related to copied media articles, while OpenAI claims that it does not have this function. On July 9th of this year, several news outlets, including The New York Times, accused OpenAI in Manhattan Federal Court of intentionally lying and misleading the court and the public regarding evidence disclosure issues.

 

CCTV reporter Xu Tao: This is outside the Federal Court in Manhattan, New York. The copyright litigation surrounding artificial intelligence training data has now extended from "whether copyrighted content has been used" to the latest stage of "whether training data can be traced and reviewed". For news organizations, the bottom line is very clear: original reports cannot be used for free to train business models; For artificial intelligence companies, the core defense is that model training belongs to the legal "fair use" and attempts to compress the public scope of data sources while protecting trade secrets.

 

Copyright lawsuit uncovers dispute over AI training data

 

In addition to this media plaintiff case, American artificial intelligence companies have also been sued for infringement by writers, film and television companies, and others. The issue of "copyright" is an unavoidable problem for AI, and there is still no clear consensus in American society on whether it is "fair use".

 

Over the years, the local news industry in the United States has been impacted by declining advertising revenue, lost subscriptions, and changes in platform distribution rules. Nowadays, if original reports are used to train commercial AI models without content producers receiving authorization income or compensation, the pressure on the local news industry will further increase. A portion of the new revenue generated by the AI industry may be built upon the long-term production of content by journalists, editors, and photojournalists. But there is still a huge controversy over whether these contents have been authorized and compensated.

 

American futurist and technology expert Ian Khan: Data is undoubtedly the fuel for the artificial intelligence economy, but human creativity is also an important source for the formation of the creator economy. Enterprises should be clear about where their training data comes from, whether it is authorized, and whether it contains protected materials or personal information. They should ensure that creators receive the rewards they deserve.

 

However, many experts say that compensating creators is technically very difficult.

 

Maribel Lopez, Chief Analyst at Lopez Research, a US technology analysis firm: They (AI companies) say they are not based on a single work, they are creating based on many works. The question is, can creators demand compensation if they want it? The answer is' no '. Because it is very difficult to determine which training data the model actually uses, you must conduct in-depth research, have professional software, and have a strong willingness.

 

In the eyes of media outlets such as The Atlantic, the copyright law in the United States was born in the era of printing presses and is fundamentally unable to adapt to artificial intelligence networks. Currently, there is no clear judicial path, and even different courts may make different judgments. Until now, the "copyright issue" in the field of artificial intelligence in the United States is still under debate.

 

Technology legal research expert Colin Levy: Innovation will continue to advance, and we cannot contain the development of new technologies. At the same time, we also need to safeguard the reasonable rights of the people. But it is obvious that the speed of technological iteration has far exceeded the pace of legal updates and improvements.

 

CCTV reporter Xu Tao: On the surface, this lawsuit is a copyright dispute, but behind it is actually a data order dispute in the AI era. Whoever masters the data will master the modeling ability; Whoever can define data usage rules may have greater say in the AI era. The problem is that the existing legal system in the United States is not fully prepared to answer these new questions.