In the "moment of drilling wood for fire" in the technological revolution, people are gradually getting used to life with AI participation. While AI writing, drawing, making short videos, and tutoring and learning, while lamenting and using AI's "superpowers", people are also worried about the corresponding risks and issues such as intellectual property rights, data security, and technical ethics.
On April 23, the country's first "AI sound infringement case" was announced, and the plaintiff was awarded 250,000 yuan in compensation; last year, the first case of AI-generated image infringement was announced, and the plaintiff was awarded 500 yuan;
The images involved in the "First AI-Generated Image Infringement Case" were originally posted by the plaintiff, Mr. Li, on his Xiaohongshu account
On the occasion of World Intellectual Property Day (April 26), we can't help but wonder: how to define the intellectual property rights of AI-generated products, and how to solve the "troubles" brought by AI to intellectual property protection while enjoying the benefits brought by AI?
1. "Take-it-or-leave-it" data "feeding" is the hardest hit area of infringement
No data, no AI. The iterative development of AI large models is inseparable from the "feeding" of large-scale data, and it is not an exaggeration to compare data to the "oil" of the information age.
"I stuff all the data into the mouth of my own AI. "The generative AI that people use most often can't do without the "feeding" of massive amounts of data if it wants to be really easy to use. While most R&D companies claim that the materials they use to train AI come from open-source libraries, copyright works and personal privacy information are often "accidentally damaged". On the one hand, there is a large AI model that is "waiting to be fed", and on the other hand, there are users who pay more and more attention to property rights protection and information protection, how to balance the two? This is not only related to data security, but also related to the vital interests of netizens.
Trik's unauthorized use of the original work as training data resulted in the software generating images that closely resemble the original
Some time ago, the statement of an office software in its privacy policy was questioned as an abuse of user privacy. After attracting attention, officials said that they had removed misleading expressions and promised that all user documents would not be used for any AI training purposes. Getty Images, an internationally renowned image website and a number of videographers and artists, has taken a series of image-generating AI developers to court, accusing them of "deliberate plagiarism and unauthorized use of millions of copyrighted images". Objectively speaking, the phenomenon of "no matter whether the user agrees or not, all data is stuffed into the mouth of the own AI" still exists.
"It's not just my work that AI steals. "The creation of AI models is not simply copying, but processing data, and even complying with the requirements of copyright law and re-editing, avoiding the comparison of "duplicate checking". However, some of the behaviors of "grafting flowers and trees and changing faces" have not been authorized by the creators themselves.
Video platform's recommendation of the AI composition tool Suno
In this way, the "transported" and "reorganized" works are disseminated and "realized" on different platforms, which not only "steals" the traffic of the original work, but also "steals" the profit margin of the creator. However, due to technical difficulties, it is difficult for people to prove that AI has captured their works and modified them, making it difficult to claim rights. At the same time, with the increasing ability of public communication, infringing activities are more scattered and difficult to prevent.
2. Does the output of AI enjoy intellectual property protection?
Alan Turing asks at the beginning of "Computing Machines and Intelligence": "Can machines think?", and Robin Li's sentence just answered him: "Artificial intelligence is not the science of the future, but the reality of the present." ”
New technologies have created more and more potential objects of intellectual property protection, including virtual character images, artificial intelligence creations, big data collections, etc., should they be protected by intellectual property rights?
Wensheng video model Sora
"When it has a 'soul', we should think about giving it ownership. "Is there copyright in the use of AI-generated content?" The judgment of the first case of AI-generated image infringement gives us the answer. If a work has "intellectual achievement" and "originality" – the "soul" of what is often referred to as an original work – then the work can be protected under copyright law.
Whether the use of artificial intelligence to generate pictures reflects the author's personalized expression needs to be judged on a case-by-case basis and cannot be generalized. For example, in the design of the picture "Spring Breeze Sends Tenderness", the plaintiff designed the presentation of the characters, selected prompt words, set relevant parameters, and continued to add prompt words and modify parameters. In this process, intellectual investment is carried out, and personal aesthetic choices and judgments are reflected, which is not a "mechanical intellectual achievement" and has originality.
"If protection is granted, who should the copyright be granted?" When the "child" of AI has a "hukou", who should be its "parents"? This question has been widely discussed by people from all walks of life. There is a view that the copyright of an AI work should belong to the programmer, while others believe that it should belong to the owner of the model.
At this stage, generative AI models do not have free will and are not legal subjects. "Essentially, it's still people who use tools to create, i.e., it's people who invest intellectually in the entire creative process, not AI models. The Beijing Internet Court made the above statement in its judgment. The general view is that the copyright of AI-generated content should belong to the entity that has contributed to the generation of the work, which may be the software developer, owner or user, which can be agreed in a contract.
3. Let AI solve problems, not become human problems
Stephen Hawking once said, "We don't need to be afraid of artificial intelligence, what we need to be afraid of is human beings." "Generative AI applications, while "smart", are not self-conscious and essentially human-developed tools for information processing.
AI is facing unprecedented intellectual property challenges, and the development of technology and the characteristics of large AI models also provide more possibilities for these illegal acts. In some cases, even legally obtained AI models can be misused, such as using the model for unauthorized applications or commercializing it in violation of licensing agreements.
How to be vigilant against people with ulterior motives and use AI to carry out deepfakes, fraud, information theft and other more covert illegal acts, "let artificial intelligence help humans solve problems, not become human problems"?
"Strengthening Information Protection" by Xinhua News Agency publisher Chunhai
"Draw a red line". AI training should keep the bottom line, and promote the "good wisdom" of AI with good governance of people. China adheres to the basic principle of people-oriented in AI governance, ensuring that the development of AI always serves humans, rather than people being controlled by machines.
In recent years, China has issued a number of regulations and policies, such as the Opinions on Strengthening the Ethical Governance of Science and Technology, the Provisions on the Administration of Deep Synthesis of Internet Information Services, and the Interim Measures for the Administration of Generative AI Services, which clearly state that generative AI should "respect intellectual property rights, business ethics, and keep trade secrets". This sets the bottom line for AI regulation of intellectual property rights - respecting intellectual property rights and avoiding infringing on the intellectual property rights of others is a requirement that generative AI services should comply with throughout the entire process from "input" to "output".
"Defeat magic with magic". The problems and challenges brought about by technology can also be solved by the progress of technology. For example, technologies such as digital watermarking, encryption, access control, and blockchain are gradually being applied to property rights protection. Just like the "thought seal" in the science fiction novel "The Three-Body Problem", these watermarks and encryption technologies have become the "subconscious" of the AI model, which can clearly mark the origin and ownership of the model at critical moments, and maintain the originality and legitimate rights and interests of the model.
Another example is the Bletchley Declaration, signed by many countries in November 2023, which mentions that AI may generate deceptive content, which may pose risks such as intentional misuse or unintentional control. The solution to the above risks also comes from data - we can solve or respond to it by obtaining good data like "gold" to train large models.
The award-winning "photos" of the German photographer are AI-generated Source: CNN
"Balancing disclosure and protection". Technology is making great strides, and the pace of governance must keep up. The intellectual property protection of artificial intelligence should be subject to inclusive and prudent supervision, which is not only a "strict father" but also a "loving mother".
Yi Jiming, a professor at Peking University Law School, once said, "Science and technology are constantly innovating, and innovation will always involve areas unknown to mankind, and when we are not yet able to fully grasp it, we should have a certain degree of inclusiveness." The supervision of intellectual property rights in the field of AI should not be too strict, so as to truly stimulate innovation.
Attaching importance to the balance between information disclosure obligations and intellectual property protection, so that AI works and human creations can have a foothold in the copyright market at the same time, compete with each other and make progress together, so as to finally achieve a beautiful picture of human-machine harmony and joint development.
On OpenAI's official website, it reads: "Let artificial intelligence create more efficient jobs for humans, and let humans benefit from artificial intelligence." "When more and more industries choose to combine with AI, and when relevant laws and regulations become more and more sound, we will usher in a newer, more standardized, and more predictable era of artificial intelligence.
(Editor: L, this article is published and produced by West Coast, from Qingdao propaganda, please indicate the source for reprinting)