Viewpoint... Analysis of intellectual property infringement risk for generative AI applications.
Published:
2023-12-14
Generating AI has certain risks when creating new content, such as copyright infringement, unfair competition, and anti-monopoly and other intellectual property infringement issues. Therefore, it is necessary to be cautious and abide by relevant laws, regulations and ethics in the process of using it.
1, The concept of generative AI
Generative artificial intelligence (Generative AI) refers to a class of artificial intelligence systems that can perform functions similar to human creativity by learning from existing data and generating new data. Generating AI systems can create new content on their own, rather than just processing and classifying it based on incoming data. These new contents may be various types of data such as text, images, audio, etc.
The core competency of generative AI is the creation of new content, not just the repeated application of known patterns. The principle is based on deep learning, especially neural network technology, which can learn the distribution and patterns of data by analyzing a large amount of training data, and then use these learned patterns to generate new data.
Generative AI has a wide range of applications, including but not limited to automatic writing and content generation, art creation and design, virtual reality and game development, scientific research and innovation, speech synthesis and music creation, education and training, medical diagnosis and image generation.
Common generative AI models include recurrent neural networks (RNNs), long and short term memory networks (LSTMs), and converter models (e. g., GPT). These models can be trained to generate new data that is similar but unique to the training data, enabling creative content generation.
It should be noted that generative AI also has certain risks when creating new content, such as copyright infringement, unfair competition, and anti-monopoly and other intellectual property infringement issues. Therefore, it is necessary to be cautious and abide by relevant laws, regulations and ethics in the process of use.
2, the generation of AI intellectual property infringement risk analysis.
2.1 copyright infringement
In the process of training and generating content, big data models may involve the risk of copyright infringement.
At the input end, the big data model needs to learn a large amount of text corpus during the construction process, which may come from the Internet, books, magazines, etc. When obtaining these text corpus, if you do not get the permission of the right holder, you may be suspected of copyright infringement. For example, an AI painting tool used images from a world-renowned image provider without permission, leading to infringement lawsuits.
On the output side, the generated content may also involve issues of copyright infringement. First of all, whether the generated content meets the requirements of "works" in the copyright law requires further judgment. If the generated content is determined to be a "work", then the issue of copyright infringement may be involved. Secondly, even if the generated content cannot meet the standard of "work", if it is substantially similar to the original work, it may still constitute intellectual property infringement.
The larger the amount of training data of the big data model, the higher the risk of intellectual property infringement of the generated content. At the same time, the user's prompt words will also affect the risk of generating content infringement of intellectual property rights. Due to the limitation of the condition setting requirements of the big data model itself, the clearer and more accurate the prompt words or instructions, the more refined the results will be, and the easier it will be to call the original corpus in its database. If the prompt is extremely limited and precise, the risk of generating content that infringes on the intellectual property rights of others increases.
2.2 infringement of trade secrets
In the training process of big data models, unknown sources or illegal data information may be used. If such information contains trade secrets, it will constitute a violation of the trade secrets of others. At the same time, with the expansion of the influence of big data model, enterprises may incorporate it into the office system to improve work efficiency. In the process of training and using the model, if employees accidentally enter the business secrets of the enterprise, it may cause the leakage and loss of trade secrets.
2.3 Unfair Competition and Antitrust Risks
Big data model has excellent performance in assisting programming, advertising design, literary creation and other fields, but there are also unfair competition and anti-monopoly risks.
Under competition law, if the user directly uses the generated content for commercial purposes, it may constitute "confusion". For example, the advertising copy generated by users using the big data model may be similar to other people's advertising copy, well-known product names, well-known enterprise names, etc., thus causing confusion.
If you use illegally crawled data for generative AI model training, the resulting data products may constitute unfair competition. Because these data products may be sufficient to substantially replace the relevant products or services provided by other operators. In addition, if a large amount of data is collected from a website with a clause prohibiting third parties from crawling data, it is likely that such data will be considered a competitive property interest.
Big data models can also raise antitrust-related risks. On the one hand, due to the high-tech nature of technology, the research and development and application of generative AI technology may become a new monopoly method for large companies; on the other hand, some companies may reach a "monopoly agreement" through artificial intelligence, such as a company through A generative artificial intelligence model analyzes the consumption habits of consumer data to adopt targeted algorithmic monopoly pricing for consumers. These actions may bring harm to consumer rights.
In order to mitigate these risks, developers and service providers need to strictly abide by relevant laws, regulations and ethics in the process of developing and using big data models. At the same time, the government and all sectors of society should also work together to formulate corresponding policies and regulations to prevent these risks.
3. Recommendations and measures
In view of the above risks, we put forward the following suggestions and measures:
Obtaining and Using Data 3.1 Legal Compliance:Ensure that the source of training data is legal and authorized by the right holder. At the same time, comply with data use agreements to avoid misuse and disclosure of data.
3.2 to strengthen the awareness of intellectual property protection:Improve employees' awareness of intellectual property protection and avoid infringing others' copyrights and trade secrets in the process of training and using models.
3.3 the establishment of intellectual property compliance mechanisms:Establish an internal intellectual property compliance mechanism, formulate intellectual property protection policies and procedures, and strengthen the supervision and management of intellectual property infringement.
3.4 enhance technical capabilities:Strengthen technology research and development and capacity enhancement, reduce dependence on large technology companies, and reduce the risk of technology monopoly.
3.5 the establishment of anti-monopoly compliance system:Understand and abide by anti-monopoly laws and regulations, avoid participating in price algorithm collusion and other monopolistic behaviors, and maintain fair market competition order.
3.6 to strengthen the protection of consumer rights and interests:Respect the rights and interests of consumers and avoid the use of large models for unfair competition and infringement of consumer rights. At the same time, strengthen communication with consumers to improve consumers' awareness and understanding of large models.
3.7 strengthen industry cooperation and supervision:Actively participate in industry cooperation and the work of regulatory agencies to jointly respond to intellectual property infringement and unfair competition risks. Through cooperation and exchanges, promote the healthy development of the industry and fair competition in society.
Key words:
Related News
Zhongcheng Qingtai Jinan Region
Address: Floor 55-57, Jinan China Resources Center, 11111 Jingshi Road, Lixia District, Jinan City, Shandong Province
