Research Paper || AI and IPR
This research paper is written by Mr. Andlib Mirza, a 3rd-year LLB student at the Campus Law Centre, Faculty of Law (Delhi University). If you also want to publish your articles or case interpretations, send your work to niyamkanoon09@gmail.com.
NAVIGATING ETHICAL AND LEGAL BOUNDARIES OF AI IN THE IP LAW FIRMS
Date: 26 June 2026ABSTRACT
Effective intellectual property rights (IPR) policy and legislation play a pivotal role in fostering an environment that is conducive to innovation and creative advancement. As technological developments continue to accelerate, the IPR framework must adapt to address emerging challenges and opportunities. One of the most significant developments in recent times is the rise of artificial intelligence (AI), which raises complex questions concerning authorship, ownership, and the scope of protection under existing legal regimes. IPR remains a fundamental pillar of the global economy, contributing not only to innovation and economic growth but also to cultural and knowledge development. In the Indian context, the evolution of intellectual property protection reflects a long and dynamic history spanning several millennia, transitioning from traditional, community-based knowledge systems to a structured, codified, and increasingly digital legal framework. The formalisation of IPR law in India can be traced to the early twentieth century, particularly with the enactment of the Indian Patents Act and Designs Act, which laid the foundation for modern intellectual property practice. In recent years, the integration of artificial intelligence into legal practice has significantly altered the functioning of IPR law firms in India. Legal professionals, particularly associates, are increasingly relying on AI-driven tools for tasks such as legal research, document review, and data analysis. While these technologies have enhanced efficiency and reduced the time required for routine processes, they also raise concerns regarding over-reliance, potential erosion of analytical skills, and the overall quality of legal work. This paper contends that although AI serves as a valuable tool in improving the productivity and operational capacity of IPR law firms, its use must be carefully regulated. There is a need to establish clear boundaries and guidelines governing the extent of AI integration to ensure that technological assistance does not compromise professional judgement, accountability, and the integrity of legal practices.
INTRODUCTION
In today’s contemporary world, it is undeniable that technological innovations have permeated nearly every facet of modern society. The depth of technology’s integration into our daily lives paints a reality where the digital realm appears to have engulfed the real world and the lines between the two are significantly blurred. The involvement of technology in the creation of innovation today is both necessary and unavoidable. The current legal system, especially the intellectual property framework, needs to evolve to align with the capabilities of AI-related technologies. The ongoing transformation demands a careful reassessment of existing laws.1 In recent years, intellectual property rights (IPR) law firms have increasingly adopted artificial intelligence (AI) tools in their day-to-day operations. These tools are now commonly used for tasks such as drafting applications, conducting trademark searches, and comparing marks. While AI has also been introduced in other areas of legal practices, particularly for case law research and document drafting, its role in IPR is notably broader and more intensive. The nature of IPR work, especially in areas like trademark prosecution, opposition, and portfolio management, often requires handling large volumes of sensitive and commercially valuable information. AI-driven tools, which rely on continuous learning and data processing, raise important concerns in this context. Unlike traditional legal tools, many AI systems improve their performance by training on the data they process. This creates a risk that confidential client information may be inadvertently stored, reused, or exposed, thereby compromising privacy and professional confidentiality. The risks are particularly significant in IPR practice because the information involved, such as brand strategies, unpublished trademarks, or business expansion plans, is often highly confidential and, if disclosed, could lead to substantial commercial harm. Clients approaching firms for trademark prosecution or opposition expect strict confidentiality, and any misuse or leakage of such data could undermine trust in legal services. Given these concerns, there is a clear need for a comprehensive framework governing the use of AI in IPR law firms. Such a framework should operate at two levels. First, law firms must develop robust internal policies that regulate how AI tools are selected, implemented, and used by professionals. These policies should include guidelines on data handling, confidentiality, and the permissible scope of AI assistance in legal work. Second, there is a need for broader national-level regulation to ensure uniform standards across the legal profession. Such policies should address issues of data protection, accountability, transparency, and ethical use of AI in legal practice.
RESEARCH OBJECTIVE
● To analyze the growing role of artificial intelligence (AI) in the functioning of IPR law firms in India
● To evaluate the impact of AI on the efficiency, accuracy, and productivity of legal professionals working in the fields of intellectual property.
● To identify the potential risks associated with the use of AI on the efficiency, accuracy, and productivity of legal professionals working in the field of intellectual property.
● To frame a human-AI collaborative model in the working of IPR firms.
RESEARCH METHODOLOGY
This research adopts a doctrinal and analytical approach to examine the growing use of artificial intelligence (AI) in intellectual property rights (IPR) law firms and the need for regulatory frameworks governing its use. The study is primarily based on secondary sources of data, including books, research articles, legal journals, policy papers, and online databases. Relevant academic literature, such as scholarly writings on the intersection of AI and intellectual property law, has been referred to in order to understand the evolving landscape and existing debates in this area.
In addition to academic sources, the research also relies on statutory provisions, policy documents, and guidelines related to data protection, confidentiality, and the use of technology in legal practice. Comparative references have been made, where relevant, to understand how different jurisdictions are approaching the regulation of AI in the legal field, particularly in the context of privacy and professional responsibility.
The methodology further includes a qualitative analysis of the risks associated with the use of AI tools in IPR practice. This involves examining issues such as data security, client confidentiality, and ethical concerns arising from the continuous learning nature of AI systems. The study also evaluates the practical implications of these risks on legal professionals and clients, especially in sensitive matters like trademark prosecution and opposition.
Moreover, the research adopts a problem-solution approach. It identifies gaps in the current legal and regulatory framework and analyses the extent to which existing laws are equipped to handle AI-related challenges in IPR practice. Based on this analysis, the study proposes the need for both internal firm-level policies and broader national regulations to ensure responsible and ethical use of AI technologies.
The overall objective of this methodology is to provide a comprehensive understanding of the subject by combining legal analysis with practical considerations, thereby offering well-reasoned suggestions for strengthening the regulatory framework governing AI in IPR law firms.
REGULATORY GAPS IN THE USE OF AI IN THE IPR LAW FIRMS
Our intellectual property laws were designed for humans like authors, inventors, and performers, not for artificial intelligence computers that can be taught on large sets of data and create outputs that look like existing copyright works. So, when an AI algorithm like Midjourney creates an image that looks like the work of Van Gogh, or ChatGPT creates text that has similarities with a copyrighted article, who is legally responsible? The programmer? The user? Or does the law leave us in a muddle with no straightforward answers? The absence of comprehensive legal frameworks and institutional guidelines governing the use of artificial intelligence (AI) in intellectual property rights (IPR) law firms poses significant risks to both clients and the judicial process. This regulatory vacuum contributes to a gradual erosion of human accountability and professional credibility, as the increasing reliance on AI tools may dilute the role of independent legal reasoning. Consequently, the potential efficiency and utility of AI remain underutilised due to the lack of clearly defined parameters for its appropriate use.
This situation necessitates the formulation of a threshold standard to determine the permissible extent of AI integration within IPR practice. Such a standard would facilitate the development of a structured human–AI collaborative model, ensuring that technological assistance complements rather than substitutes professional judgement. In addition, there is a pressing need to establish clear liability mechanisms for legal professionals who employ AI tools, particularly in instances where errors, data breaches, or professional negligence arise.
The issue assumes greater significance in the context of client confidentiality. When a client approaches an IPR law firm for services such as trademark registration, there is a legitimate expectation that the matter will be handled with the utmost care, discretion, and confidentiality. The disclosure or misuse of sensitive information, including brand strategies or unpublished marks, may result in irreparable commercial harm. This concern has led several IPR law firms to adopt a cautious approach, with some opting to restrict or prohibit the use of AI tools in their operations.
However, a complete prohibition on AI is neither practical nor desirable. The challenge lies in ensuring its appropriate and proportionate use. Much like any tool, AI must be applied within its intended scope and limitations. Its effectiveness depends not on unrestricted deployment but on calibrated use supported by adequate safeguards. When integrated within a system of checks and balances, AI can significantly enhance efficiency in legal processes, including IP litigation, without compromising professional standards.
Ultimately, the central issue remains unresolved: the question of attribution of responsibility in AI-assisted legal work. In the absence of clear legal standards, determining liability for errors or misconduct arising from AI use continues to be a complex and unsettled matter. Addressing this gap is essential for ensuring accountability, maintaining client trust, and upholding the integrity of the legal profession.
FRAMEWORK FOR AI GOVERNANCE IN VARIOUS LAW FIRMS OF INDIA:
- ANAND AND ANAND : This law firm has integrated specialised expertise in artificial intelligence within their intellectual property (IP) practice, enabling them to develop strategies that are aligned with emerging technological advancements. This integration allows firms to better assist clients in maximising the commercial value of AI-driven intellectual property while navigating the associated legal complexities. The presence of professionals with specialised knowledge in AI within IP law firms helps mitigate potential risks arising from the use of such technologies. Their expertise ensures that client work is handled with greater caution, particularly in areas involving data sensitivity, authorship concerns, and regulatory compliance. Additionally, these professionals play a crucial role in facilitating the creation and protection of AI-generated intellectual property by proactively identifying and addressing legal and ethical risks associated with such innovations. Furthermore, such specialists contribute to the responsible use of AI tools within legal practice by guiding their application in a controlled and informed manner. This includes ensuring that inputs and interactions with AI systems are appropriately structured so as to maintain accuracy, confidentiality, and reliability in legal outputs.
- KHAITAN AND CO. : One notable development in this context is the creation of an in-house artificial intelligence tool, “Khaitan & Co AI” (KAI), designed to enhance efficiency in traditionally time-intensive legal processes. Such an institutional framework represents a significant step towards establishing a structured and controlled model for AI integration within intellectual property (IP) practice. By enabling the automation of routine and research-heavy tasks, it allows legal professionals to redirect their focus towards higher-value strategic and advisory functions. Importantly, this system is designed with a strong emphasis on data security and client confidentiality. It operates through integration with external large language models via Microsoft Azure, while ensuring that all data is appropriately redacted and not permanently stored. In addition, the firm is in the process of developing a proprietary small language model hosted entirely within its internal infrastructure, thereby maximising data privacy and minimising external exposure. A key feature of this framework is its robust governance structure. Access to sensitive information is strictly regulated, and permissions are dynamically managed such that when a team member disengages from a particular matter, their access to related documents is automatically revoked. This reflects a conscious effort to maintain accountability and prevent unauthorised data exposure. Further, all AI-generated outputs are clearly labelled with visible disclaimers, such as “AI-generated – Kindly review", thereby acknowledging the possibility of inaccuracies and reinforcing the necessity of human oversight. This mechanism serves as an important safeguard against issues such as algorithmic errors or "hallucinations" while preserving professional responsibility. The firm has also undertaken regular training initiatives for its legal professionals, including both associates and partners, to ensure awareness of compliance obligations, data privacy standards, and practices such as the redaction of personally identifiable information (PII). Such initiatives promote a comprehensive human–AI collaborative model, bridging the gap between technologically adept younger professionals and those accustomed to traditional methods of legal practice. In doing so, the framework ensures inclusive and effective adoption of AI across all levels of the organisation.
INTERNATIONAL FRAMEWORK AND SUGGESTIONS FOR THE WORKING OF IPR ADVOCATES AND FIRMS
- WORLD INTELLECTUAL PROPERTY ORGANISATION: The World Intellectual Property Organization’s (WIPO) initiative, the WIPO Conversation on Intellectual Property and Frontier Technologies, serves as a significant global platform for dialogue and knowledge exchange among stakeholders on the impact of emerging technologies, particularly artificial intelligence, on intellectual property (IP) systems. The eighth session of this forum specifically focused on the intersection of generative AI and IP, with the objective of assisting policymakers in understanding and evaluating potential regulatory approaches. In its publication titled Generative AI: Navigating Intellectual Property, WIPO provides a comprehensive analysis of generative AI, including its functioning, the concept of confidential information, and the various risks associated with its use. The publication highlights critical concerns such as the potential for generative AI systems to compromise confidentiality and contribute to intellectual property infringement. One of the primary risks identified is that generative AI tools may retain and utilise user inputs for training purposes. Consequently, if users include confidential or sensitive information within prompts, such data may be stored by the service provider, leading to a loss of confidentiality and possible misuse. Alongside identifying these risks, the publication also proposes practical solutions to mitigate them. For instance, it recommends that users carefully review and adjust the settings of generative AI tools to minimise the likelihood of data retention or training on user inputs. It further suggests the adoption of AI systems that operate within secure environments, such as private cloud infrastructures, to enhance data protection and confidentiality. Additionally, the publication outlines a set of measures aimed at promoting the responsible and legally compliant use of AI by businesses and organisations. These measures are presented in the form of a checklist intended to guide stakeholders in implementing safeguards against potential legal and ethical risks associated with generative AI. Following are the few measures highlighted:
- Implement a staff policy and training to guide appropriate usage and to encourage responsible experimentation and use of generative AI
- Regularly assess and update policies based on evolving risks and court decisions.
- Communicate legal risks clearly to the business to adopt practices according to the business risk appetite.
- Ask staff to label AI-generated output and to keep records of prompts used.
- Document the role of humans in the creation process.
- Vet datasets when training AI and consider IP ownership and licence coverage
- Integrate human input and creativity with AI outputs to maintain control over ownership of outputs.
- Check for IP infringements before using outputs.
- EPI GUIDELINES: These guidelines (adopted at the C98 Council on 16 November 2024) have been prepared by the Professional Conduct Committee of EPI to assist members who use generative AI in their work as European patent attorneys. The overarching principle of these guidelines is that when using AI of any kind in professional work, a member must adopt the highest possible standards of probity, must take all reasonable steps to maintain confidentiality when this is required, and at all times must put the interests of the clients first. Following are the guidelines:
- Guideline 1: Members must be adequately informed about the general functioning of generative AI and the specific risks of the models they use, particularly concerning confidentiality and hallucinations.
- Guideline 2a: Members must ensure appropriate confidentiality of all data shared with AI systems and refrain from using such tools where confidentiality cannot be guaranteed.
- Guideline 2b: Members must understand the potential modes and risks of unintended disclosure associated with specific AI models.
- Guideline 3a: Members remain fully accountable for their professional work and cannot rely on AI use as a defence for errors or omissions.
- Guideline 3b: Members must rigorously review AI-generated outputs to ensure they meet the standard expected of competent human practitioners.
- Guideline 4: Members must ascertain and respect client consent before using generative AI in any professional engagement.
- Guideline 5a: Any disclosure regarding the use of AI in professional work must be accurate, fair, and non-discriminatory.
- Guideline 5b: Disclosure of AI use in formal proceedings is not mandatory unless required by law or client instruction but must remain accurate and appropriate where made.
- Guideline 6: Members must adopt safeguards, such as separate user accounts, to prevent cross-client data exposure where confidentiality risks exist.
- Guideline 7a: Members must comply with all applicable legal and regulatory provisions governing the use of AI.
- Guideline 7b: Members must also consider and adhere to external organisational obligations or restrictions relating to AI use.
- Guideline 8: Fees for AI-assisted work must be fair and proportionate to the effort, complexity, and risk involved, including costs related to AI tools and verification processes.
EFFICIENT HUMAN-AI COLLABORATIVE MODEL
In view of the issues discussed above, it is necessary to critically examine the different models of human–AI collaboration with the objective of identifying the approach that ensures optimal efficiency, accountability, and reliability in the functioning of intellectual property rights (IPR) law firms.
- 100% AI-DRIVEN MODEL: This represents the apex of technological autonomy, relying on fully autonomous algorithmic agents to execute end-to-end tasks with no human intervention, oversight, or midstream correction. In this framework, the AI system functions as the primary, and solely active, reasoning element and decision-maker.3 Empirical data from a landmark 2025 study conducted by researchers at Stanford University and Carnegie Mellon University demonstrates the raw volumetric power of this approach. The study revealed that fully autonomous agents required 88.3% less time to complete knowledge-intensive tasks than human workers operating independently. Furthermore, these agents utilised 96.4% fewer actions to reach a conclusion and reduced overall operational costs by an astonishing 90% to 96%. A study conducted by Stanford University's Human-Centred Artificial Intelligence (HAI) institute found that general-purpose AI tools hallucinated between 58% and 82% of the time on complex legal queries, rendering them highly dangerous when operating without human guardrails. The US Copyright Act of 1976 and high-profile international rulings, such as the DABUS patent cases, have repeatedly affirmed that an AI system cannot be legally recognised as an inventor or an author. Consequently, any creative work, legal draft, or patent application generated entirely by a 100% AI-driven model is categorically excluded from intellectual property protection, meaning the resulting asset immediately falls into the public domain. The 100% AI-driven model is wholly unsuitable for final patent drafting, legal reasoning, bespoke strategic advice, and definitive infringement detection, where algorithmic hallucinations present catastrophic professional liability risks.
- 50-50 HUMAN-AI COLLABORATION MODEL: This model is predicated on a “shared-control” dynamic, aiming for an equilibrium where human and artificial agents possess roughly equal decision-making authority and cognitive load. This model envisions a continuous, synchronised partnership rather than a sequential handover of tasks. This model aligns with the concept of cooperative AI, where agents are capable of autonomously performing helpful tasks while remaining continuously aligned with human preferences and intent. A primary advantage is real-time risk mitigation. Continuous human involvement allows for the immediate identification and correction of statistical anomalies, hallucinations, or logic failures before they are embedded into the final legal product. A demonstrative industry case study highlights this dynamic in utility management: during a severe winter storm, an AI system (GridAssistant) continuously flagged potential transmission line faults and suggested load redistributions. A human operator, utilising years of experience, instantly validated and authorised these suggestions, enabling the AI to implement them in milliseconds. Despite its theoretical elegance, achieving a genuine 50-50 balance is practically difficult and introduces significant workflow friction. Traditional patent law struggles to allocate rewards in 50:50 scenarios, raising unresolved legal questions about co-authorship versus derivative works and potentially jeopardising the enforceability of the resulting intellectual property.
- AI DOMINANT HYBRID MODEL: The AI-dominant hybrid model inverts the traditional professional dynamic. In this framework, the algorithmic system is elevated to the role of primary researcher, initial drafter, and core decision-maker, while the human's role is relegated to the periphery typically functioning as a downstream reviewer, editor, or passive approver of machine-generated outputs. The primary advantage of the AI-dominant hybrid is its ability to facilitate high-volume data processing while maintaining a nominal level of human oversight. This model is exceptionally adept at handling massive datasets where human intervention is only required for edge cases. By automating routine document review, early-stage patent classification, and basic summarisation, the AI-dominant model can reclaim vast amounts of time estimated at nearly 200,000 hours annually for a large enterprise law firm. The AI-dominant model harbours significant hidden costs and behavioural vulnerabilities, most notably "automation bias" and the "debugging penalty". When humans are positioned strictly as downstream reviewers, psychological studies demonstrate a strong tendency to succumb to automation bias accepting AI outputs without adequate critical scrutiny because the machine appears authoritative. Behavioural performance evaluations indicate that over-reliance on AI significantly reduces human critical thinking, causing the AI-dominant scenario to rank lower in overall decision quality compared to human-dominant hybrid setups.
- HUMAN-DOMINANT HYBRID MODEL: The human-dominant hybrid model is characterised by "human-led workflows augmented by AI". In this paradigm, the human practitioner firmly retains total strategic control, defining the parameters, initiating iterative prompts, and holding sole decision-making authority while utilising the AI as an incredibly powerful cognitive prosthesis. This model yields the highest Cognitive Sustainability Index (CSI), fostering cognitively rich engagement, self-reflection, and continuous learning on the part of the human practitioner. It capitalises on the mathematical concept of "complementarity", where the combined performance of a human and an AI exceeds either acting in isolation. Because the human and the AI possess low latent correlation in their error patterns (they make different types of mistakes), their combined effort drastically reduces overall systemic error. Crucially, the human-dominant model ensures absolute ethical and legal defensibility. By keeping the human firmly in the loop as the primary creator and decision-maker, this model preserves legal accountability, maintains attorney-client privilege, and explicitly ensures that the resulting work meets the "human spark" threshold required for copyright and patent protection under WIPO and global regulatory standards. The primary limitation of this model is that it is highly dependent on specialised human skill sets. To achieve complementarity, the human operator must possess an accurate mental model of the AI's specific capabilities, strengths, and blind spots. Practitioners must undergo rigorous upskilling to master advanced "prompt engineering" and develop the critical capacity to assess AI outputs; otherwise, the technology is severely underutilised.
CONCLUSION
The future of intellectual property rights in the algorithmic era is not defined by a zero-sum competition between human intellect and computational power but rather by the deliberate, highly structured collaboration between the two. The empirical evidence is unequivocal: a race toward 100% autonomous AI deployment in complex legal tasks results in unacceptable error rates, profound ethical breaches, and the ultimate nullification of intellectual property protection due to the absence of human authorship. Conversely, strict adherence to a 100% human-driven model ensures legal reliability but courts systemic obsolescence, economic inefficiency, and a failure to scale justice in the face of exponential data growth. The human-dominant hybrid model, where strategic human foresight and ethical judgement are systematically augmented by AI's unprecedented data processing capabilities, represents the optimal global standard. By carefully allocating tasks according to the task-technology fit, instituting rigorous human-in-the-loop validation workflows, and adhering to strict WIPO transparency guidelines, organisations can harness the transformative efficiency of artificial intelligence while completely neutralising its inherent risks. Ultimately, the successful IP practitioners of tomorrow will not be those who ignore AI, nor those who surrender their professional judgement to it; rather, they will be the experts who master AI as a collaborative prosthesis, elevating the practice of intellectual property law to unprecedented heights of strategic and operational excellence.
