Title: Exploring the Compliance Framework for Business Development
Author: May Pang, ChainDeDe
Author's Bio: Head of Legal Affairs at Oort, expert in law, risk, and compliance, with extensive experience in top financial institutions such as Standard & Poor's and Bank of America, as well as start-ups in the Fintech and Web3 industries.
The implementation of Artificial Intelligence (AI) is rapidly expanding across various industries, especially with generative AI technology dominating the headlines in 2023. Consequently, regulators and practitioners are increasingly concerned about the application of AI technology and its implications for intellectual property (IP) and personal data privacy.
Recently, I attended a legal summit in New York where industry professionals engaged in heated discussions about the legal and compliance issues surrounding AI technology. It was widely acknowledged that as AI technology continues to evolve, companies need to consider the following factors when formulating their AI strategies.
1. Patent and Copyright Issues for AI-generated Content
Earlier this year, the AI copyright lawsuit filed by computer scientist and entrepreneur Stephen Thaler garnered widespread media attention. Thaler used AI technology to develop the DABUS system and began applying for patents globally. However, to date, patent offices in the US, EU, UK, Australia, and New Zealand have rejected his applications, except for South Africa. The reason behind this is the consensus among many intellectual property regulatory bodies that only content created with human involvement is eligible for copyright protection, and the inventor eligible for copyright protection must be the original designer of the copyrighted content. AI-generated content does not meet the criteria for copyright protection.
In March 2023, the US Copyright Office issued guidelines on copyright determination for AI-generated content and further sought public opinion in August 2023 on whether legislation should be enacted in this area. A tip for AI practitioners is that if they want copyright protection, their AI works must include sufficient human-created content, rather than being entirely AI-generated. When applying for copyright protection from the US Copyright Office, creators must also disclose how much of the content in their works is generated by AI. Therefore, companies involved in AI-generated content should establish more detailed internal processes and maintain detailed records of the content creation process for future copyright applications.
2. Fair Use Exemption
Currently, most AI products are built on accessing large amounts of data to train AI models. For example, the data used by OpenAI possesses characteristics such as "publicly accessible, third-party licensed, and user-generated." However, publicly accessible data may also contain copyrighted content. This is where the "fair use" exemption principle often comes into play.
The "fair use" exemption allows users to use copyrighted works in certain limited circumstances without payment, including quoting original text for commentary, criticism, writing news reports, and academic reports. Generally, the "fair use" exemption principle provides protection for data users. However, it is important to note that if the original work is altered or transformed into a new form during the use, the "fair use" exemption principle may become invalid.
The application of the "fair use" exemption varies by jurisdiction, so practitioners need to understand the specific legal requirements in each operating location. The US judicial system currently has no definitive guidance on whether the "fair use" principle universally applies to AI products. Different companies have varying responses to the "fair use" principle. OpenAI insists that their use of publicly available information should be protected by the "fair use" principle. On the other hand, Stability AI, a company using AI to generate images, faced a lawsuit in the US ("Stability AI case") due to its lack of consideration for the copyright issues of original artists when obtaining data from the internet. Although in a hearing in July 2023, the presiding judge ruled that there was no "substantial similarity" between the images created by the plaintiff artist and those generated by Stability AI's AI system, this serves as a warning for all AI companies that collect public data. Companies using public data to train their AI products should be cautious to prevent substantial similarity between their generated products and any works used as input for their AI training data to avoid future litigation.
The Federal Trade Commission (FTC) recently released best practice guidelines suggesting that to prevent public suspicion of intellectual property violations, companies providing generative AI products should proactively disclose the copyrighted materials included in their AI training data to enhance the transparency of their products. In 2021, the United Nations also advocated for increased transparency in product creation in its recommendations on AI ethics, which has since become a core aspect of many government policies on AI.
It is worth noting that the EU and Australia have proposed another approach to copyright protection for AI data, which is to provide a "selective opt-out" mechanism for companies holding patents or intellectual property. This means that copyright holders can explicitly indicate that they do not want their patents or intellectual property to be cited or used by others, giving companies that do not want their data to be used for training AI systems the right to choose. However, this practice has also faced challenges in implementation because even if your image is later removed from the AI database, it is currently unclear how AI can "forget" these removed images after learning.
3. Data Licensing
The news industry has been at the forefront of data licensing. News Group reportedly has been in discussions with AI companies on how their published news content can be used by AI companies for training their models for a fee. The Associated Press has also reached an agreement with OpenAI to share their content and technology, and explore potential cooperation in the AI field. These initiatives indicate that data providers have the opportunity to collaborate with AI companies to mutually address licensing issues. However, some media outlets, such as The New York Times, CNN, and Disney, have taken a more stringent stance in response to content licensing, prohibiting GPTBot (OpenAI's data scraper) from extracting information from their content. Although no lawsuits have been filed yet, The New York Times is considering legal action against OpenAI. This tense relationship underscores the urgency and necessity of dialogue between both parties.
4. Data Privacy Protection
As the data used to train AI may contain personal information, strengthening data privacy protection is a pressing issue for companies in the AI industry. For example, in March 2023, the video conferencing company Zoom quietly modified its terms of service to grant itself the right to use customer data for AI training, which sparked widespread concerns among users. Zoom revoked the modification to its terms of service a few days later and clarified that it would not use any user content for AI training.
Currently, one approach that AI companies are taking to address data privacy issues is data anonymization, which involves removing "sensitive" information from the data, such as personal banking information or medical records, while retaining other basic data. This enhanced privacy data is widely used in many jurisdictions worldwide. In the EU, companies may be required to inform users about the potential use of their personal information through a privacy statement before processing the data, thereby proactively increasing transparency. The state of Delaware in the US has also begun to develop its own consumer privacy laws.
Conclusion
In summary, AI technology is rapidly advancing and significantly impacting various industries. Regulators and practitioners are actively exploring how to achieve business development within a compliant framework, taking more measures to protect the intellectual property of creators, promote greater transparency in the operations of AI companies, and engage in cooperative dialogues between data providers and product developers. This field is full of opportunities and challenges, and the better prepared one is, the more likely they are to succeed in future competition.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
