How should enterprises establish compliance frameworks amid the AI wave?
TechFlow Selected TechFlow Selected
How should enterprises establish compliance frameworks amid the AI wave?
Regulators and practitioners are actively exploring how to achieve business development within a compliant framework.
Navigating Web3 tides with focused insightsBy May Pang, TechFlow
Author Bio: Legal, risk, and compliance expert at Oort, with extensive experience in top financial institutions such as S&P and Bank of America, as well as startups in the Fintech and Web3 industries.
The implementation of artificial intelligence (AI) is rapidly expanding across industries, particularly generative AI technologies that dominated headlines throughout 2023. As a result, regulators and practitioners are increasingly focused on legal and ethical issues arising from AI applications—especially concerning intellectual property (IP) rights and personal data privacy protection.
Recently, I attended a legal summit in New York where industry professionals engaged in vigorous discussions about the legal and compliance challenges posed by AI technology. There was broad consensus among participants that as AI continues to evolve, companies developing their own AI strategies must carefully consider several key factors.
1. Patent and Copyright Issues for AI-Generated Content
Earlier this year, computer scientist and entrepreneur Stephen Thaler’s AI copyright lawsuit made headlines worldwide. Thaler developed an AI system called DABUS and filed patent applications for it globally. However, except in South Africa, patent offices in the United States, the European Union, the UK, Australia, and New Zealand have all rejected his applications. The primary reason lies in prevailing views among intellectual property regulators: only content created with human involvement qualifies for copyright protection; inventors protected under copyright law must be the original creators of the work. AI-generated content does not meet current criteria for copyright eligibility.
In March 2023, the U.S. Copyright Office issued guidance on copyright registration for AI-generated works and further solicited public comments in August 2023 on whether new legislation should be introduced for this domain. A key takeaway for AI practitioners is that to qualify for copyright protection, their AI-assisted creations must include sufficient human authorship—not be entirely generated by AI. When applying for copyright registration with the U.S. Copyright Office, creators are also required to disclose how much of the work was produced by AI. Therefore, companies involved in AI-generated content should establish internal processes and maintain detailed records of content creation workflows to support future copyright claims.
2. Fair Use Exemption
Most current AI products rely on access to vast datasets to train models. For example, OpenAI uses data characterized as “publicly accessible, third-party licensed, and user-generated.” However, even publicly available data may contain copyrighted material—this is where the doctrine of “fair use” exemption often comes into play.
“Fair use” allows limited use of copyrighted works without permission or payment under specific circumstances, such as quoting excerpts for commentary, criticism, news reporting, or academic research. Typically, fair use provides some legal protection for data users. However, it's important to note that when copyrighted material is significantly altered or transformed during use, the applicability of fair use may no longer hold.
Fair use standards vary by jurisdiction, so practitioners must understand local regulations in each region they operate. The U.S. judicial system has yet to issue definitive guidance on whether fair use generally applies to AI training practices. Companies currently take different positions. OpenAI maintains that its use of publicly available information falls within fair use protections. In contrast, Stability AI, an AI image generation company, became involved in a U.S. lawsuit (the "Stability AI case") due to allegations of insufficient consideration for original artists’ copyrights when sourcing web data. Although in a July 2023 hearing the presiding judge ruled there was no “substantial similarity” between plaintiff artists’ works and images generated by Stability AI’s system, the case remains a cautionary tale. Companies using public data to train AI systems must ensure their outputs do not bear substantial resemblance to any input works used in training, to avoid future litigation risks.
The U.S. Federal Trade Commission (FTC) recently released best practice guidelines recommending that generative AI providers proactively disclose if their training data includes copyrighted materials, aiming to increase transparency and prevent public suspicion of IP violations. In 2021, the United Nations also recommended corporate transparency in AI development as part of its AI ethics guidance. Since then, transparency in AI creation has become a core principle in policymaking across many governments.
Notably, the EU and Australia have proposed an alternative approach to AI data usage: an “opt-out” mechanism allowing rights holders to explicitly prohibit the use of their copyrighted materials. This gives companies and creators who do not wish their data to be used in AI training the right to object. However, practical challenges remain. Even if an image is removed from an AI database, it is unclear whether—and how—the AI model can truly “unlearn” previously ingested data.
3. Data Licensing
In terms of data licensing, the news industry has taken an early lead. News Corp is reportedly in discussions with AI companies over compensation for using its published articles to train AI models. The Associated Press has entered into an agreement with OpenAI to share content and technology and explore potential future collaborations in AI applications relevant to both parties. These initiatives show that data providers can collaborate with AI developers to resolve licensing issues through mutually beneficial arrangements. On the other hand, some media organizations—including The New York Times, CNN, and Disney—have adopted a stricter stance, blocking GPTBot (OpenAI’s web crawler) from accessing their content. While no lawsuits have been filed yet, The New York Times has indicated it may pursue legal action against OpenAI. This tension underscores the urgency and necessity for constructive dialogue between content owners and AI developers.
4. Data Privacy Protection
Because data used to train AI systems may contain personal information, strengthening data privacy protection has become an urgent priority for AI companies. For instance, in March 2023, video conferencing company Zoom quietly updated its terms of service to claim rights to use customer data for AI training—a move that sparked widespread user concern. Within days, Zoom reversed the change and clarified it would not use any user content for AI training purposes.
One common method AI companies use to address privacy concerns is data anonymization—removing “sensitive” identifiers such as personal banking details or medical records while retaining non-sensitive data. Such privacy-enhanced data processing methods are now widely adopted across many jurisdictions. In the EU, companies may be required to inform users via privacy notices before processing personal data, thereby increasing transparency. Delaware in the U.S. has also begun enacting its own consumer privacy laws.
Conclusion
In short, AI technology is advancing rapidly and profoundly impacting various fields. Regulators and industry players are actively exploring ways to drive innovation within compliant frameworks—by better protecting creators’ IP rights, promoting greater transparency in AI operations, and fostering collaborative dialogues between data providers and product developers. This space presents both immense opportunities and significant challenges. The earlier companies prepare, the more likely they are to succeed in the competitive landscape ahead.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
