The Profound Influence of Advanced AI Technology on the Legal and Intellectual Property Sectors and Strategic Response
date: 2024-04-24 Samson G Yu, Lucy Li Read by:

The advent of generative artificial intelligence (AIGC) products, such as ChatGPT at the close of 2022, has accelerated the evolution of AI at a staggering rate. In under a year, we've witnessed a transition from the chat and text generation capabilities of ChatGPT 3.5 to the superior performance of GPT-4. More recently, just two months ago, we saw the introduction of SORA, capable of generating videos with a single command. Simultaneously, the influence of AI on the legal services sector is deepening and broadening rapidly, presenting an urgent challenge for the industry. Leading entities, like the esteemed law firm Allen & Overy, have pioneered the adoption of AI innovations, having developed an intelligent legal assistant in partnership with Harvey Company in the United States. Meanwhile, many in the field are contemplating which strategic paths to pursue in response to these advancements


1. The Impact of AIGC on the Legal and Intellectual Property Industry


1.1. Beyond Digitization and Informationization


The transformative power of AIGC stems from its core functionality as a generative model that leverages advanced machine learning, particularly deep learning technologies, to craft novel content such as text, images, music, code, and 3D models. This technology transcends traditional digitization and informatization—which mainly translate real-world data into manageable digital formats for analysis. AIGC transcends mere conversion or translation of real-world inputs. It exhibits increasingly human-like capacities to learn, comprehend, generalize, and innovate, utilizing these capabilities to generate unprecedented forms of content.


For example, ChatGPT4 has already acquired the ability to interpret images and generate textual descriptions. Below are examples of a patent application process diagram and ChatGPT4's automatic interpretation in an animated format.



1.2. AIGC and Human Characteristics


AIGC, built on expansive language models and extensive data training, possesses the capability to swiftly process and analyze vast datasets, producing detailed content. This technology exhibits human-like qualities such as learning, summarizing, generalizing, and analytical reasoning. As a silicon-based intelligence, it can operate continuously around the clock, supported by robust computational power. In contrast, humans, as carbon-based intelligence, might not match AIGC in terms of efficiency, stamina, and data handling. However, the depth and complexity of human creativity remain unparalleled and more nuanced.


1.3. The Legal Industry as a Suitable Field for Deep, Vertical Application of AI

In the legal sector, tasks such as providing consultancy, drafting documents, appearing in court hearings, and conducting negotiations are commonplace. This industry is prolific in generating a wide array of documents, including case analyses, contracts, legal opinions, and research reports, all of which serve as crucial references or deliverables. Consequently, the legal sector can be viewed as a content-generation industry, one that traditionally involves significant manual effort due to the complexity and variety of case contexts. However, with AIGC systems trained on diverse datasets, there is a capacity to automate the creation of various legal documents by learning from extensive repositories of legal regulations and case histories. This positions the legal field as a prime example of a vertical market where AI can be applied most effectively.


The patent domain, characterized by its unique blend of legal and technical elements, presents significant opportunities for deploying AI tools. For example, consider the following two animated demonstrations: an AI large model adeptly summarizes the content of the Supreme Court’s Patent Infringement Judgment (2010) No. 158. This summary includes detailed interpretations such as the validity of the claims, the substantiation of infringement facts, and an analysis of the technical innovations in prior art, specifically AU2009270482. Notably, the AI excels in precisely extracting and elucidating innovation points from distinct sections of the specifications.



1.4. Traditional Law Firms' AI Innovation and the Rapid Development of Startup Intelligent Legal Service Providers


Recognizing the transformative impact of generative artificial intelligence on the legal sector, prominent law firms like Allen swiftly formed partnerships with Harvey Company. Leveraging its specialized legal databases for pre-training and fine-tuning, Harvey equipped Allen’s attorneys with a bespoke ChatGPT assistant. This tool aids in drafting documents, conducting legal research, reviewing contracts, explaining terms, and generating pertinent recommendations. By the end of March, another major global law firm announced its own initiative with Harvey to integrate generative AI technologies.


This evolution has catalyzed the emergence of new categories of legal AI enterprises. Firstly, there are companies like EVEN UP, which provide AI-driven legal services online, specializing in areas such as personal injury claims and achieving multi-billion dollar valuations. Secondly, there are firms like Harvey, which utilize models like ChatGPT, and Robin AI, which employ ANTHROPIC’s Claude technology, to develop tailored AI solutions for businesses and law firms. These companies have experienced rapid growth, evidenced by significant increases in funding, client bases, and revenue. For instance, Harvey Company’s annual recurring revenue escalated from a few million dollars in April 2023 to tens of millions by December 2023.


1.5. The Rapid Development of AI Platforms and Tools Greatly Expands the Space for Practical Applications


As a large language model, ChatGPT-3.5 has been able to perform tasks such as drafting letters, translation, and text grammar modification excellently. These tasks are usually handled by legal assistants, but ChatGPT-3.5 can easily complete them with a single click. Its generated business letters are thoughtful and polite, requiring little to no further modification. In terms of translation, it can achieve a fairly accurate level. By inputtingrigid documents into ChatGPT, it can immediately adjust sentences and correct grammar, transforming them into perfect manuscripts.


Beyond the conventional duties of legal assistants, large language models like ChatGPT have made notable advancements in drafting legal documents and research reports. Within a year, ChatGPT evolved from version 3.5 to 4.0, with the latter demonstrating a comprehensive grasp of legal principles and the ability to autonomously outline various legal documents, such as warning letters, complaints, and contracts. With ongoing input and refinement from professional lawyers, GPT-4.0 can produce increasingly detailed legal documents, examples of which are illustrated in the animated demonstrations provided. These frameworks enable lawyers to tailor the final documents to specific case scenarios.


In technology-related legal fields, such as patent law, where there is an intersection with technological and industrial sectors, large models are not only well-fed with extensive training data but also adept at navigating technical literature, extracting key information, responding to technical queries, forecasting technological trends, and even drafting competent technical research reports. Additionally, innovative tools have emerged that can generate patent application documents from technological disclosures or draft specifications directly from claims, like Patentpal, though these products still have potential for enhancement.


For instance, the animations below show how GPT-4 generates payment terms for a patent license contract, and, following iterative refinements through lawyer interactions, adds terms to enhance the oversight of licensing revenues.


Over the past year, the realm of commercial large language models has seen significant advancements, marked by the evolution of ChatGPT into GPT-4. Additionally, a range of distinctive AI platforms developed by companies such as Google and ANTHROPIC have surfaced, some of which outperform GPT-4 in specific domains. For instance, one particular model demonstrates remarkable accuracy in answering technical questions, as illustrated in the animation below, which precisely describes the latest developments in PAO catalyst preparation and identifies the most promising catalyst types.


Recently, the launch of the GPT Store marked a major milestone, with over 3 million applications now available, spanning a broad spectrum of office and legal professional tools. Innovations like Midjourney and Sora are pushing the boundaries of large language models into the multimodal realm, enabling the generation of images, audio, and video. Moreover, the dynamic and complex landscape of open-source large models is providing numerous developers with opportunities to create and customize private large models, further enriching the AI ecosystem.


1.6. AI Narrows the Gap in Professional Knowledge and Tools, Highlighting Personalized Traits


Whether proprietary or open-source, AIGC's expansive knowledge base plays a crucial role in progressively bridging the knowledge gap among professionals who utilize similar AI applications. Furthermore, the broad array of platforms and tools offered by AIGC applications equips diverse professionals, further diminishing disparities in knowledge. However, this convergence of expertise is not guaranteed, and the extent and nature of this knowledge gap reduction will differ based on the specific characteristics of organizations or individuals involved.


In the legal domain, the use of GPT tools facilitates rapid access to a wealth of professional legal knowledge. As demonstrated in the animations below, various GPT applications provide insights into U.S. patent law, Information Disclosure Statements (IDS), and Inter Parties Review (IPR), showcasing how these tools can effectively respond to specialized inquiries in the field.



1.7. What Kind of Legal Work Will Be Replaced by AI?


From current trends, it is clear that aspects of legal work that solely entail the provision of knowledge are poised for automation. The mere recollection and reproduction of such knowledge can be fully substituted by AI technologies, while the nuanced selection of knowledge in legal contexts will not be entirely supplanted. Legal services that rely on replicable knowledge and can be systematized are beginning to see automation, starting with augmentation. However, tasks that require creative and personalized legal expertise will prove more challenging to automate. These areas, dependent on human ingenuity and tailored advice, remain less vulnerable to replacement by current AI capabilities.


2. Countermeasures for Law Firms under AIGC


Faced with the arrival of the era of artificial intelligence, firms providing general legal and intellectual property and other specialized legal services should actively think, adapt to changes, and take action to respond to the transformation brought about by artificial intelligence.


2.1. Developing Enterprise AI Response Strategies


The rapid advancement of AIGC has broadened its application scope and deepened its influence, introducing a multitude of new products that are reshaping the legal services industry. This transformation is impacting business models, skill requirements, service content, job roles, and team structures in profound ways. From the vantage point of legal service firms, there is a tangible possibility that legal professionals might be supplanted to some degree by AI agents. This shift could disrupt traditional industry paradigms, positioning AI companies as pivotal partners and suppliers for legal enterprises. As a result, professionals within this sector will increasingly require AI-driven empowerment to stay current with technological updates and iterations.


In response to evolving circumstances, enterprises must craft comprehensive AI-centric strategies alongside their conventional business development plans. These strategies should establish clear objectives centered on the integration and application of AI technologies and products. Detailed plans should be laid out for foundational AI technology development, innovative service projects, process enhancement, and the cultivation of personnel capabilities. Additionally, these strategies must address ethical considerations and risk management, ensuring robust financial and human resource allocations to support these initiatives effectively.


2.2. Building Enterprise-Specific Large Models


With the rise of open-source large model projects, the feasibility of developing customized large models (LLMs) for enterprises has become increasingly apparent. In the legal services industry, firms should weigh considerations such as data governance, competitive edge, and cost efficiency when building bespoke large models. These tailored models can significantly enhance the quality of AI tool utilization by addressing gaps in professional knowledge and industry-specific data accumulation present in generic models, thereby enhancing the specificity and accuracy of outputs. They also help meet stringent data governance standards, safeguarding against the leakage of sensitive information and ensuring compliance with the data protection requirements of clients, governments, and other stakeholders. Enterprises can host these private large models on local servers, training them on relevant data to generate confidential and proprietary results. Essential elements of private large models include:

Element 1: Data Construction The uniqueness of a private large model stems from the data it trains on, which includes both industry-specific data like legal regulations, litigation documents, and patent information, and proprietary enterprise data such as internal case files and knowledge bases. Enterprises must first undertake digital transformation projects to convert their data into a format suitable for training large models. Those with advanced IT and digital capabilities in the legal field will be at an advantage in this new era of model training.

Element 2: Model Selection

Various open-source large models have continually emerged, each distinguished primarily by differences in architecture and algorithms. Leveraging these open-source models, enterprises can construct their own tailored large models by training data and applying fine-tuning to develop private models suited to their specific needs. These models vary in architecture, number of parameters, volume of training data, and capability features. Enterprises in the legal services sector should thoughtfully select models based on a comprehensive evaluation of model reliability, security, and licensing compliance to meet their anticipated goals.


Element 3: Computational Support Training and deploying large models require substantial computing resources, which can lead to significant infrastructure costs. Comparable to the way humans need food, AI needs computational power. Whether through cloud services or local servers, the computational demands are considerable—as evidenced by companies like NVIDIA, dubbed the "King of AI Chips," which achieved a market valuation of 2 trillion USD due to the high demand for AI GPUs, especially evident in markets like China where GPU shortages are common.

Element 4: Team Building Creating a private large model necessitates a collaborative effort between AI specialists and legal industry experts. AI professionals handle the technical aspects of model development, such as customizing parameters and managing operations, while legal professionals provide expert knowledge, assist in selecting and training data, and participate in the AI training process. Firms can develop these capabilities internally or through partnerships with external organizations.



2.3. Enhancing the Application of Commercial AI Platforms and Software


The deployment of commercial AI platforms and software plays a crucial role in the modern technological landscape. These platforms, typically closed-source, frequently provide capabilities that surpass those of their open-source counterparts. Nevertheless, it is essential to exercise caution by not inputting sensitive and confidential information into these platforms, thereby safeguarding data security.

Complementary and Verification Among Multiple Platforms

Beyond ChatGPT, other significant commercial AI platforms have been developed by entities such as Google and ANTHROPIC, marking the emergence of a competitive market dominated by several large models. These platforms exhibit unique strengths—some excel in processing extended texts and offering multimodal functionalities, while others are superior in conversational abilities or boast more rounded capabilities. The diversity in their technical foundations and training methodologies has led to varied competencies, which can be leveraged and validated through cross-platform applications.

Exploring the Application of General and Professional AI Tools

The evolution of commercial large models into comprehensive platforms has solidified their role in legal applications. For instance, ChatGPT has transcended its original function to serve as a platform akin to an App Store. With the recent launch of the GPT Store, these models have showcased their robust platform potential. The store's initial offerings include 3 million applications spanning seven categories—image generation, writing, office productivity, research and analysis, programming/software development, education, and lifestyle. This breadth and customization capability promise a continuous expansion in the number and variety of applications, enriching the tools available for the legal sector.


Many of these tools are directly applicable to legal services. Writing tools like All Around Writer, Human GPT, and Copywriter GPT can produce a range of professional documents and marketing content. Productivity tools such as Slide Maker and Canva, along with programming and software development tools, allow for the creation of customized software solutions tailored to the specific needs of the legal industry.


Multimodal Large Model Application


SORA’s ability to effortlessly generate videos depicting Tokyo street scenes has captured widespread attention. In the legal industry, the predominant use of text is well established, whereas the application of images, audio, and video remains relatively underutilized. However, every law firm operates not just as a provider of legal services, but also as a business entity that must engage in branding and create relevant multimedia content. Therefore, enhancing the use of multimodal large models in these aspects is critical.


2.4. Developing "Intelligent Digital Employees" - AI Agents


With advancements in AIGC technology, the concept of "intelligent digital employees" has become increasingly viable. These AI agents, powered by large language models, autonomously plan and execute tasks, effectively replacing certain human work functions. Unlike traditional automation software or virtual avatars, these agents understand, plan, and implement tasks through sophisticated AIGC capabilities. By developing such AI agents, routine and transactional tasks can be delegated to digital employees, allowing human lawyers to focus on client relationships, deep client needs analysis, judgment-making, and creative tasks.


Setting AI Agent Employee Positions and Capability Requirements


Legal service enterprises should identify roles for AI agents based on their business processes and define the capabilities required of these digital employees. Roles could range from customer service agents addressing common inquiries to secretarial agents managing case workflows, and even AI assistants helping lawyers with drafting, billing, and legal research. Additionally, there's potential for AI agents to handle brand promotion and multimedia content creation.


Exploring Management Models for AI Agent Employees


Integrating AI agents poses challenges in balancing functions between human employees and digital counterparts and achieving synergistic outcomes. Managing these intelligent digital employees effectively, particularly in detecting and responding to unexpected problems and risks, will be crucial.


2.5. Service Project Innovation and Process Optimization


The adoption of AI tools, private large models, and intelligent digital employees is transforming traditional legal services. This evolution not only changes operational processes but also introduces new service offerings.


Service Project Innovation


With private large models at their disposal, legal firms may offer services directly utilizing these models, with competition likely to focus on the accuracy and efficiency of these models in routine legal tasks. In more complex legal matters, the emphasis may shift towards delivering high-value content and distinguishing between high and low-value tasks, using AI to alleviate lawyers from mundane activities, thereby enhancing quality and reducing costs.


Service Process Optimization


Service process innovation in legal firms will primarily involve automating routine tasks such as contract reviews, document drafting, and data organization through AI. Communication with clients and collaboration between AI agents and human staff should aim for prompt responsiveness and seamless information exchange. Quality control and efficiency assessments will increasingly rely on AI-driven data analysis to monitor progress and productivity.


2.6. Personnel Team Organization and Capability Building


In the AI-driven era, legal service enterprises must actively enhance team organization and develop capabilities to not only raise AI awareness among team members but also to significantly boost their professional knowledge and proficiency in applying AI technologies.


Team members ought to have a foundational understanding and practical skills in AI and associated technologies, utilizing tools like ChatGPT, Gemini, and Claude. They should explore how AI can be integrated across different operational aspects of the business and proactively incorporate these tools into daily work practices. Given the rapid evolution of AI technology, a commitment to lifelong learning is essential to keep pace with the latest advancements and applications. Moreover, team members must be capable of assessing the reliability and logical consistency of outputs from AI systems, and should be skilled in analyzing and interpreting these results to extract valuable professional insights and strategies.


As AI becomes increasingly integral to legal practice, it is crucial for employees to have a strong sense of ethical and legal responsibility. They must be adept at discerning which tasks are suitable for AI solutions and identifying any potential ethical or legal implications of the AI-generated outcomes.


Regarding job design, enterprises need to transcend traditional structures and introduce roles like legal knowledge engineers and prompt engineers, who will be tasked with data management and training of large models. This approach aims to leverage AI tools to their fullest potential, fostering the creation of private large models and ensuring meticulous oversight of day-to-day activities within the firm. This strategic deployment of AI tools will enable firms to offer innovative suggestions for service models, content, and process enhancements, driving continual improvement and adaptation in the legal services sector.


3. Risk Management for Legal Service Enterprises under AIGC


Enterprises that integrate artificial intelligence technology must rigorously manage the associated risks. This is especially critical for legal service enterprises, where the stakes in managing these risks are particularly high. Here are key risk areas:

3.1. Risk of Harmful Outputs

Artificial intelligence systems, powered by large language models, operate differently from human understanding. They analyze vast amounts of data to identify language patterns and predict subsequent words, which is fundamentally different from human comprehension of language and context. Consequently, if the training data contains biases, these systems may generate outputs that inadvertently perpetuate or amplify these biases. The risk escalates with the use of non-proprietary models, where lack of transparency in training data can lead to unexpected and potentially harmful outputs.

3.2. Data Security Risk

Using both commercial AI tools and non-proprietary models involves significant data security concerns. Information input into these models can be stored and potentially disseminated online, exposing sensitive and confidential enterprise data to risks such as leaks to competitors or breaches by malicious actors. Furthermore, this compromised data could be utilized to train new model versions, inadvertently converting proprietary knowledge into publicly accessible information. This situation poses a direct challenge to meeting the data compliance standards expected by clients, governments, and other relevant entities. Even with private large models, robust management practices are essential to mitigate the risk of data leakage.

3.3. Legal and Ethical Risks

Current AI technologies, including large language models, lack the capability to truly understand context, discern truth, or uphold justice. Their outputs can be inaccurate, potentially discriminatory, and not aligned with ethical norms. Overreliance on AI in legal contexts can lead to deviations from the core mission of legal practice, producing advice or opinions that may compromise fairness and justice. It is imperative for legal service providers to enhance their oversight and management of AI tools to ensure they uphold the legal, ethical, and moral integrity of their services. This includes careful evaluation and refinement of AI-generated content to align with professional standards and societal values.



4. Future Outlook

The swift evolution of artificial intelligence technology, along with its products and commercial applications, marks a profound shift in the legal industry, signaling an inevitable integration into its core operations. Legal service institutions and practitioners must therefore align with these industry and technological advancements, taking proactive steps to navigate the challenges and harness the opportunities presented by this new era of artificial intelligence.


Legal service enterprises play a central role in this transformation. At the awareness level, they must courageously depart from traditional work and business models, showing a readiness to commit substantial effort and resources to the integration of artificial intelligence. Practically, they should formulate and implement comprehensive strategies to enhance their utilization and management of AI technologies, identifying and capitalizing on the new business opportunities these technologies bring. Internally, the adoption of AI tools can significantly boost work efficiency and reduce labor costs through the development of AI-agent employees. Externally, this entails a reinvention of business and commercial models, exploring new service areas and innovative projects.


Legal service practitioners are at the forefront of implementing these adaptive measures. They must embrace continual learning to expand their knowledge and capabilities, breaking away from the legal profession’s historical resistance to innovation. By actively integrating artificial intelligence tools, they can improve the efficiency and quality of their work, adapt to new operational models, and significantly enhance their professional value in the era of artificial intelligence.