2024 PSSA Digital Policy Award Runner-up: Samyukta Srinivasan

---

Striking a Balance

Navigating Intellectual Property Law and Innovation in Generative AI for the Canadian Legal Landscape

By Samyukta Srinivasan

Introduction 

The dynamic landscape of Artificial Intelligence (AI) technology is becoming increasingly difficult to navigate, particularly with the rise of generative AI, which is a type of AI technology capable of creating new content that is derived from some input content, and is able to mirror the complexity of this input data. Unlike traditional AI which relies solely on pre-existing data, generative AI has the ability to create new content, which is similar but not identical to original data, based on deep-learning models trained on text, audio, visual and other complex data types (Government of Canada, 2024).

While undoubtedly a revolutionary development, this type of technology has had significant implications for innovation, copyright issues, and the ethical use of AI technologies. It stands at the forefront of AI research and application, pushing the boundaries of creativity and machine learning capabilities. Intellectual Property (IP) has been defined as the “socially, culturally, and economically useful products of the human intellect” (Leung, n.d.). IP rights constitute rights that grant the holder power to legally stop others from using their IP without their consent (Leung, n.d.). Generative AI has resulted in the infringement of IP rights, and has threatened rights of use issues, uncertainty about ownership of AI-generated works, and has posed questions regarding unlicensed content in training data, and whether users should be able to prompt these tools with reference to other creators’ original works without their knowledge (Vargas, 2022). Even more perplexing is the question: How should we proceed when IP is generated directly by AI (Olijnyk, 2022)?

The Problem

AI has had notable impacts on the commercial interests of creators, artists, and authors as it trains on vast troves of data, including copyrighted materials, thereby infringing upon the rights of original content creators (ITALIE, 2023). Even though Canadian copyright law protects original works of authorship, the emergence of AI as a ‘creator’ blurs the lines of legal definitions and ownership (CCH Canadian Ltd. V. Law Society of Upper Canada, 2004). For example, AI-generated art such as that produced by OpenAI’s DALL-E-2, blurs the lines of ownership and originality, potentially undercutting traditional creators’ business and royalties. Moreover, the lack of transparency from AI creators regarding data sourcing and usage further complicates matters. On the commercial front, without any embargo on commercial use of these AI generated images, original creators of art can only hope to lose more business in terms of volumes as also royalties due to availability of cheap alternatives for their art generated by AI.

Thus, Canada’s existing legal system must adapt as the advent of machine learning and new AI technologies continue to challenge the very definitions of ownership and human ownership that once defined the country’s legal framework for the determination of IP rights.

Evaluating Existing Regulations & Policy Gaps 

The introduction (and ongoing consideration) of Bill C-27, also known as the Digital Charter Implementation Act 2022 (DCIA), shows that a cohesive (albeit still vague) nation-wide regulatory framework is in the works. If brought into force, Bill C-27 will go on to solidify the proposed Artificial Intelligence and Data Act (AIDA), a legislative endeavor which outlines the responsibilities of individuals and legal entities “responsible” for AI systems; this includes those who design develop or provide AI systems or manage their operations during the course of international or interprovincial trade and commerce (Medeiros & Beatson, 2022). Essentially, AIDA introduces regulatory measures for “high-impact” AI systems, focusing on risk mitigation, transparency, and prohibition of practices that could cause “material harm” (Medeiros & Beatson, 2022).

After its publication, experts highlighted several inconsistencies and discrepancies embedded within the text of the AIDA. One concern highlighted was the Act’s ambiguity; for example, the proposed legislation fails to clearly define which systems fall under the category of “high-impact” and what constitutes “material harm” (Medeiros & Beatson, 2022). This type of ambiguity is bound to complicate the interpretation and enforcement of the law, and could potentially subject researchers, businesses, and private individuals to substantial penalties. Furthermore, a lack of specific definitions within AIDA could potentially enable Innovation, Science and Economic Development Canada to implement expansive AI regulations without open and transparent public deliberation.

Recommendations 

Based on the gaps that have already been identified in AI laws in Canada, the development of clear legal definitions and guidelines for AI-related matters would be an important starting point. For example, the Canadian Copyright Act could incorporate specific definitions and provisions for AI-generated works, distinguishing between AI-assisted and AI-generated content, to clarify the status of works produced with the significant intervention of AI systems (Innovation, Science and Economic Development Canada, 2023). Furthermore, certain legal measures must also be aimed at AI developers; for instance, AI developers can be mandated by the law to implement a system for tracking the use of copyrighted material in training datasets and ensuring fair compensation to original creators. Another measure aimed at AI developers could be to disclose the datasets they use to train their AI systems, especially when these systems are utilized for commercial purposes. This disclosure should be made without compromising data privacy or proprietary information, possible through a regulatory body overseeing AI development and use. These, of course, are long-term solutions that will require approximately 2-3 years to actually come into force as any changes to patent or copyright law have certain time lags associated with

them (Huber, 2023). However, the incorporation of AI in the patent system is something that must be considered, particularly due to the unprecedented rate at which AI continues to evolve and the line between content generated by machine learning and real human beings continues to become blurred.

Another key recommendation is to create a centralized digital licensing platform overseen by a dedicated state or central body, where creators can list their works available for AI training under predetermined terms and conditions. From this platform, AI developers will be able to obtain the necessary licenses to streamline the process and ensure legal clarity. Additionally, the establishment of a compliance monitoring body with the authority to audit and enforce the use of copyrighted materials in AI is also an important step towards ensuring adherence to licensing agreements. These measures will enable copyright holders, AI developers, and legal experts to work towards developing fair remuneration models that reflect the value contributed by copyright works to AI training datasets through the implementation of fixed fees, revenue sharing models, or usage-based payments.

It is also important to establish and promote clear ethical standards for AI use and development practices, and ensure accountability for businesses that do not comply with these standards. For instance, collaborating with regulatory bodies such as the Standards Council of Canada for the development of ethical standards for AI use in content creation. Businesses that are certified (i.e. that meet these standards) would signal compliance with ethical use of IP and respect for copyright laws. Another way to enhance and promote ethical standards could be through public awareness campaigns delivered through workshops or online resources, which could serve to educate both AI developers and the general public on the importance of ethical AI development and its impact on IP rights. Additionally, encouraging stakeholder engagement through ongoing dialogue between AI developers, IP owners, and policymakers through forums and roundtables, could aid in continuously refining ethical guidelines to keep pace with the rapidly-evolving nature of AI technologies.

Finally, fostering collaboration between AI developers and creators is crucial for bridging the gap between IP law and innovation. This could be facilitated by the establishment of innovation incubators focused on collaborations between AI developers and IP creators. To incentivize developers and creators to collaborate, the Government could offer tax incentives and grants to projects that demonstrate effective collaboration, by focusing on the development of AI technologies that respect and enhance the value of IP. Furthermore, such collaborative relationships can only exist and complement one another by ensuring transparency. This could be brought about by investing in the research and development of blockchain technology for IP management, such as creating blockchain-based systems for tracking and managing IP rights and transactions in a transparent and secure manner (FasterCapital, 2017).

Conclusion 

The rapid advancement of generative AI technologies present challenges and opportunities for IP laws in Canada. As we stand at the precipice of a new era in content creation, innovation, and digital rights management, it is imperative that Canada’s legislative framework evolves in tandem with these technological strides. The interplay between AI-generated content and IP rights necessitates a nuanced approach to regulation – one that fosters innovation while protecting creators’ rights. Policymakers, industry leaders, and the academic community must collaborate to refine Canada’s AI governance model, ensuring it remains adaptable, equitable, and effective in the face of AI’s evolving landscape.

---