Bar Association of Erie County Bulletin
October/November 2024
Alexander C. Kaczmarek
Last time, Cyber Law reviewed the impact, challenges, and future potential of generative AI patent drafting and prior art search programs be defining the claimed abilities of such programs and reviewing President Biden’s Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence and the United States Patent and Trademark Office’s guidance on use. Undoubtedly, the emergence of generative AI tailored towards patent drafting has great potential; however, the use of generative AI area also presents unique risks that practitioners and organizations must carefully consider. This article explores the practical risks and benefits of using generative AI in patent drafting, providing a balanced perspective for patent practitioners.
Benefits of Using Generative AI in Patent Drafting
- Increased Efficiency
Generative AI tools can significantly accelerate the patent drafting process. Traditional patent drafting is lengthy, often requiring multiple rounds of revisions and consultations. Generative AI programs can automate much of this process by drafting initial and revised applications, and performing prior art searches. This increased efficiency allows patent attorneys to focus on more complex or strategic aspects of patent prosecution, such as claim strategy and office actions.
2. Increased Efficiency
One of the most compelling advantages (for law firms and clients alike) of using generative AI is cost reduction. By automating routine tasks, generative AI programs can reduce the time spent on drafting patents, which in turn lowers billable hours for clients, or lower discounts in receivables for law firms. This can make patent services more accessible to smaller businesses and startups, who may otherwise be deterred by high legal costs, and increased billing efficiency for law firms and practitioners.
3. Consistency
Generative AI adheres to predefined templates and language models. This can reduce the likelihood of errors that might arise from human oversight, such as inconsistent terminology or misclassification of inventions. Generative AI programs can also help predict possible Office Actions, based on those raised in similarly structured applications and prior art, with an aim of reducing the risk of an application being rejected on such grounds.
4. Enhanced Innovation
Finally, generative AI programs can arguably assist in identifying aspects of an invention that might not be immediately apparent to a practitioner or inventor. By analyzing the claimed invention against large datasets of prior patents and technical literature, generative AI programs can suggest different ways of describing an invention, potentially uncovering broader or more robust claims. This can result in stronger patents that provide more comprehensive protection for an invention.
However, in weighing these benefits, practitioners must be aware and mindful of the risks.
Risks of Using Generative AI in Patent Drafting
- Quality and Precision Concerns
While generative AI programs can draft patents quickly, there is a risk that the quality and precision of these drafts may not meet the standards of experienced patent attorneys. Patent drafting is a nuanced process that requires a deep understanding of the invention, the relevant legal framework, and strategic foresight. Generative AI programs may struggle with these complexities, leading to drafts that are too generic, overly broad, or insufficiently detailed to withstand scrutiny during examination or litigation. While generative AI programs may introduce efficiency and consistency in its initial and revised drafts, the time required for a senior attorney’s review and edits to such drafts may surpass the time spent on a traditionally drafted application.
2. Scope of the Invention
Further, while generative AI programs may assist in efficient patent drafting, it may fail to actually capture aspects of the invention that are valuable to the client and their business. The complexity of claim drafting is not always defining the invention in its entirety, but tailoring the claims to those aspects of the invention that the client specifically wishes to protect or are likely to be infringed. In this way, the prior benefit of cost savings can be drastically impacted. For example, a narrow patent claim may obtain a quick allowance, but a generative AI program may draft a broad set of claims that require multiple responses which could wipeout any cost savings in drafting.
3. AI Hallucinations
Use of generative AI programs also introduce significant risk in the potential for AI “hallucinations,” where the program generates information or content that is either inaccurate or entirely fabricated. For example, a generative AI program might produce technical descriptions, prior art references, or legal arguments that appear valid on the surface but are not grounded in reality. These hallucinations can be particularly dangerous, as they may go unnoticed by a human reviewer who relies too heavily on the generative AI output (especially where the human reviewer is a newer practitioner or the invention is in a complex technical area), leading to flawed patent applications. Such errors could result in significant setbacks, including rejections, disputes, or the issuance of weak patents that fail to provide adequate protection. Therefore, it is crucial for practitioners to remain highly vigilant and thorough when reviewing generative AI outputs for accuracy and relevance.
4. Bias and Limited Understanding
Finally, on risks towards the quality of generative AI drafts, generative AI programs are trained on an unknown-to-practitioners set of existing data, which may include biases or gaps in knowledge. This can lead to generative AI outputs that reflect outdated practices or fail to account for emerging trends and technologies. Since generative AI programs rely on existing data and patentable inventions are, by definition, novel, generative AI programs may struggle to understand the unique aspects of highly specialized or cutting-edge inventions. This risk could result in significant setbacks, including rejections, disputes, or the issuance of weak patents that fail to provide adequate protection.
5. Confidentiality and Data Security
Patent applications often involve sensitive, confidential information about new inventions. Using generative AI programs that operate in the cloud or rely on external servers raises concerns about data security. If the generative AI program is compromised, this could lead to the exposure of trade secrets or other proprietary information not otherwise included in an application. Further, where a generative AI program reuses a practitioner’s inputs for training, a client’s proprietary information may filter into outputs provided by the program to other third-party users. Not only would this be a potential breach of a practitioner’s ethical duty to maintain confidentiality of information, but such disclosure and reuse of the existence of and enabling information on the invention could potentially be considered a non-confidential public disclosure triggering the start of a grace period for filing a patent application in the United States and barring the filing of an application in some other countries that require complete novelty as a requirement for patentability. Thus, practitioners must understand the security measures and licensing requirements of the generative AI program at issue.
6. Over-Reliance on AI
As the use of generative AI programs becomes more widespread, there is a risk that practitioners may become overly reliant on use of such tools, leading to a degradation of their own drafting skills over time. This over-reliance could also result in more relaxed approach in reviewing and editing generative AI outputs, increasing the risk of errors or oversights. It is crucial for practitioners to maintain a balanced approach in the introduction and use of generative AI programs, using generative AI programs as a tool to complement, rather than replace, their expertise.
7. Replacement of Junior Associates and Limiting Development
The risk of over-reliance additionally impacts and poses a potential threat to the traditional training and development of junior associates in law firms. Traditionally, junior associates gain experience by drafting patent applications, conducting prior art searches, and reviewing and incorporating the edits of senior attorneys in future drafts. As generative AI programs are used for more of these foundational tasks, there is a risk that junior associates may have fewer opportunities engaging with these tasks, limiting their professional growth. This could also lead to a future generation of practitioners who are less equipped to handle complex patent issues, particularly those that require deep legal insight and creativity that would be developed through time working closely with senior attorneys. Finally, the reduced need for junior associates in certain tasks might lead to fewer entry-level opportunities in the field, potentially limiting the pipeline of new talent entering the profession.
Despite the cost saving potential of generative AI programs, law firms dedicate significant resources into training and development of junior associates by billing junior associates at a lower rate to account for any extra time required and, in certain cases, providing clients with discounts on junior associate time to account for such training and development. In these circumstances, the cost of a patent application may be similar or the same, but the client loses the benefit of a trusted advisor that is developing a specialized knowledge in that client’s subject matter in favor of an “all-knowing” generalist (being the generative AI program). Law firms must carefully balance the integration of generative AI programs with the need to maintain robust training and development programs to ensure that the next generation of patent professionals is well-prepared to meet the challenges of the field.
Conclusion
Generative AI holds great promise for improving the efficiency, cost-effectiveness, and consistency of patent drafting. However, in its current form, it is not without significant risks, and practitioners must weigh the benefits of use of generative AI programs against these risks. Ultimately, the most effective use of generative AI programs in patent drafting will involve a hybrid approach, where human expertise and generative AI program capabilities are combined and, as the technology evolves, so too will the strategies for integrating generative AI into the patent drafting process.