In this written interview with FORUM Institute, I shared my personal experiences and insights on AI-Assisted Patent Drafting. Here are the questions asked:
Question 1: Martin, you have been dealing with AI-generated patent drafting for a while. When have you seen the real potential for making the lifes of patent attorneys easier? It would be great for the readers of the FORUM IP Talks to get an overview of the development of automated patent drafting.
From Early Experiments to Modern Solutions
Yes, I have seen the evolution of AI-assisted patent drafting right under our noses. 13 years ago, I started programming a simple patent drafting tool based on Visual Basic and Excel. There are still traces from that time, such as a joint patent publication with TU Darmstadt. Although these tools were rudimentary, they laid the foundation for the more sophisticated automation technologies that followed.
The Impact of Transformer Technology
A major turning point came in 2019 with the introduction of transformer technology, which significantly improved the quality of text generation. We have organized the 1st World Robot Patent Drafting Symposium in the summer of 2022, and we have again left traces: there is a corresponding whitebook available on Amazon that documents the status quo at that time. The Large Language Models (LLMs) marked a new era in efficiency and accuracy, allowing for more complex and nuanced language generation in patent drafting.
Revolutionizing Patent Drafting with ChatGPT
The release of ChatGPT in late 2022 a short time after our Symposium further revolutionized the field, making advanced AI-driven patent drafting capabilities widely accessible. With this development, AI tools moved from niche applications to becoming an integral part of mainstream patent drafting processes.
Key Features of AI-Assisted Patent Tools
Today, AI tools have changed the patent drafting landscape by automating various aspects of the process. These systems can generate detailed figure descriptions from figures, support patent claims with precise benefits and technical effects, and retrieve background information such as technical definitions of claim features. They also excel at specialized functions such as translating patent applications and creating software patent flowcharts from patent claims. The benefits of these advances are profound. Tasks that once took hours or even days can now be completed in minutes. AI ensures consistency in terminology and technical descriptions, greatly reducing the risk of errors. Anyone who has ever written a patent application knows how frustrating it is to be interrupted by a phone call. These days are gone: new patent drafting tools provide the flexibility to allow interruptions in the drafting process without causing major setbacks, as structured workflows can be seamlessly resumed.
Future Innovations in AI for Patent Drafting
In the future, AI tools will continue to improve. They will increase the speed and accuracy of patent drafting. I am currently developing a solution to automate the creation of a support language for the patent claims. This support language helps to incorporate even arbitrary combinations of technical features from the description into the claims after filing a patent application, if necessary, addressing the European Patent Office (EPO) requirements for “intermediate generalizations”. In addition, better methods for converting U.S.-style patent applications into litigation-grade patents suitable for Unified Patent Court (UPC) litigation are expected soon. These are just two of the many improvements to come.
Question 2: Recently at the autumn conference of the German association of IP experts, you spoke with Sebastian Goebel on the application of generative AI in IP. In a post here on LinkedIn, Sebastian quoted from a conversation between two AI legal tech experts: “AI will not replace attorneys, but attorneys with AI will replace attorneys without AI.” What do you think, would they apply to patent attorneys in patent practice and also to corporate patent professionals?
AI and the Patent Profession
It is a fact that the synergy between human expertise and machine efficiency is central in the field of patent drafting. By integrating advanced AI tools with strategic legal guidance and the specialized technical expertise of patent professionals, the industry is well positioned to achieve faster innovation and higher productivity.
More importantly, the quality of the AI-generated output is highly dependent on the quality of the input. This highlights the need for humans to capture and refine invention details before being converted into a patent application by a robot.
The Synergy of AI and Human Expertise
Early developers of AI-driven patent drafting tools believed that these systems would eliminate the need for patent attorneys, or even for humans in general. However, experience shows that inventors often struggle to describe their inventions in sufficient detail without the guidance of a trained patent professional. The same applies to interpreting prior art documents in the light of a patent claim: while identifying literal matches are easy to find for a Large Language Model LLM, identifying equivalent, implicit, or inherent disclosures of features in a prior art document is still a challenge for AI-assisted software. As a result of poor input data, AI tools can neither generate a coherent description of the embodiments of an invention nor produce satisfactory patent claims.
Although AI excels at generating readable text, it lacks the ability to make wise decisions. For this reason, the statement “AI will not replace attorneys, but attorneys with AI will replace attorneys without AI” resonates deeply in the fields of patent practice and corporate patent management.
AI’s Role in Transforming Patent Attorneys’ Workflows
As a result, for patent attorneys, incorporating generative AI is about more than following technology trends; it is about transforming how they efficiently deliver high-quality work. Generative AI streamlines labor-intensive tasks, such as creating lengthy descriptions of technical figures or supporting claim features with technical effects and advantages taken from sources on the web. By automating these processes, patent attorneys can spend more time on strategic tasks, such as optimizing patent claim scope and preparing the patent application to become a litigation-grade patent. Patent attorneys who integrate AI into their workflow will gain a competitive advantage over their peers who rely solely on last century’s methods. The situation of patent attorneys today is like that one of taxi companies 100 years ago: the introduction of the internal combustion engine did not make taxi companies obsolete. It was just that a cab company with motorized cabs did better than one that stuck with its traditional horse-drawn carts.
Generative AI in Corporate IP Management
The same principle applies to corporate patent professionals, who are responsible for managing large portfolios and overseeing innovation pipelines. Generative AI enhances their ability to perform tasks such as invention harvesting, prior art searches, patentability assessments, and freedom-to-operate analyses with much greater speed. In-house IP professionals using AI tools can provide timely, data-driven advice to their organizations, giving them a significant advantage by accelerating the innovation cycle.
AI Cannot Replace Human Judgment
It is important to emphasize that AI is not a replacement for human expertise. AI serves as a powerful augmentation tool, providing efficiency, consistency, and analytical insights that complement human capabilities.
Staying Competitive with AI Integration
In summary, the effective use of AI will likely distinguish forward-thinking patent attorneys and corporate IP professionals from those who are slower to adapt. Those who integrate AI into their practice will be better equipped to navigate the increasingly competitive and fast-paced IP landscape, confirming the enduring relevance of the statement, “AI will not replace attorneys, but attorneys with AI will replace attorneys without AI.”
Question 3: What are the common fears when it comes to using generative AI for patent drafting, and how do you think they can be addressed?
Addressing Fears of AI in Patent Drafting
I speak from experience because I have trained more than 200 patent professionals in the use of AI-assisted patent drafting tools: most participants expressed at least one of four common concerns about using AI tools:
- • Fear of data theft
- • Fear of data leakage
- • Fear of creating unwanted prior art
- • Fear of compromising client confidentiality
Understanding Common Concerns
Client confidentiality remains the critical concern mainly for patent attorneys who are using AI tools because for those – different from their inhouse peers – there are strict legal regulations in place that can turn client confidentiality issues into a serious criminal matter.
The integration of generative AI into patent drafting has also raised several common fears, particularly around intellectual property theft when uploading invention descriptions to a service on the Internet and the risk of data leakage in the form of prompt data used to train patent drafting robots. After rationalizing these fears, the solution is to address these concerns with technical safeguards and operational practices to ensure the responsible and effective use of AI-assisted drafting bots.
Ensuring Data Security and Confidentiality
A related concern is the risk of novelty-damaging prepublication. Information submitted to free or inadequately secured AI systems could potentially be exposed, inadvertently entering the public domain and compromising the novelty of an invention. To prevent this, patent professionals should avoid using publicly available AI tools. Instead, they should use enterprise solutions with robust legal agreements that explicitly prohibit data storage or sharing by the AI provider.
Maintaining Control Over the Drafting Process
A less common fear is the potential loss of control over the drafting process. Many users are concerned that AI-generated drafts will be too generic or fail to reflect the nuanced judgment required for high-quality patent applications. To address this, AI must be viewed as an assistant, not a replacement. AI-generated drafts should always be treated as initial suggestions, subject to thorough review, adaptation, and refinement by experienced patent professionals who can ensure technical and legal accuracy.
Managing AI Bias and Error Risks
Concerns about bias and error in AI-generated content are not common. These risks arise because AI models may inherit biases from their training data or struggle to effectively handle highly technical or niche subject matter. To address this, it is common sense that users must ensure that the input provided to the AI is well-structured and specific to guide the system toward accurate outputs. It is not necessary to mention that any AI-generated content must be rigorously validated by experts before being included in a final patent application.
Complementing, Not Replacing, Human Expertise
Many people worry about losing their jobs due to AI, even if they do not say it openly. Some professionals are concerned that AI tools may reduce the need for human patent attorneys and corporate IP specialists. This fear comes from direct experience with AI-assisted patent drafting software. For example, it is possible to create a 20-page patent application with 15 claims in just 2 hours using AI. The same task would take 10 or more hours using traditional methods.
My own experience as a patent attorney with 28 years of professional experience together with the experience of having trained more than 200 patent professionals in the use of AI-assisted patent drafting tools comes in handy when putting that perceived speed gain into perspective. First, the initially “decently looking” patent applications always turn out to be not good after more careful inspection. Human input is needed to turn them into a patent application that can be filed with good consciousness. This is real measurement data: the average speed gain when drafting patent applications with AI-assisted patent drafting software is 50%, and not 90% as one might expect after the initial jaw-dropping experience. In a side-line only: according to my experience, about 20% of the users of AI-assisted patent drafting software experience a speed gain of 3 times or 4 times, as compared with traditional patent drafting methods. But there are about 20% of the users of AI-assisted patent drafting software that do not experience any speed gain, as compared with traditional patent drafting methods, half of which are even slower than using traditional patent drafting methods. I am cautious to mention that fact in public when discussing job-loss fears, but it is a truth backed up by enough observations.
In practise, the job-loss concern can be mitigated by emphasizing the complementary nature of AI. Rather than replacing human expertise, AI increases efficiency and accuracy, allowing patent professionals to focus on the strategic, creative, and high-value aspects of their work that cannot be automated.
In summary, while generative AI in patent drafting raises legitimate concerns, these challenges can be effectively addressed with thoughtful measures. By selecting safe and reliable AI tools, implementing robust safeguards, closely monitoring AI results, and embracing the role of AI as a complement rather than a replacement, patent professionals can confidently harness the benefits of AI while managing its risks. This balanced approach allows professionals to realize the full potential of AI while ensuring the integrity and quality of their work.
Question 4: How can concerns, especially about security and confidentiality, be addressed in detail? For some potential users of generative AI among the patent people, this seems to still represent quite a hurdle.
Overcoming Security and Confidentiality Challenges
Addressing security and confidentiality concerns is fundamental to building trust in the use of generative AI for patent drafting. Patent data is extremely sensitive, and any mishandling by AI systems could lead to significant risks. Implementing a multi-pronged approach can effectively mitigate these concerns and ensure the safe use of AI tools in professional workflows.
The Importance of Data Security in Patent Drafting
First, users should prioritize AI tools from vendors that enforce strong security measures, such as end-to-end encryption and clear contractual confidentiality guarantees. In industries with particularly sensitive data, on-premises AI solutions that operate entirely within the company’s infrastructure provide an additional layer of security and control.
AI tools from reputable vendors with a demonstrated commitment to data security adhere to industry standards by employing measures such as end-to-end encryption, two-factor authentication (2FA), and other access controls; ensuring that data is encrypted in transit and at rest; and anonymizing data to remove identifying details. Robust contractual agreements are also critical, explicitly guaranteeing confidentiality and prohibiting the provider from storing or using user input data beyond immediate processing. Advanced technical measures, such as role-based access permissions help limit unwanted access to sensitive data. In addition, firewalls and intrusion detection systems protect internal networks from unauthorized access, enhancing overall security. These commercial AI platforms also comply with relevant privacy laws, such as the General Data Protection Regulation (GDPR) in Europe or the California Consumer Privacy Act (CCPA) in the United States. Compliance demonstrates a commitment to high privacy and security standards and provides additional peace of mind for users.
Leveraging On-Premises Solutions for Maximum Safety
On-premises AI solutions operate entirely within an organization’s internal network, ensuring that sensitive data never leaves the premises. While this option may be more expensive, it provides complete control over data. It minimizes exposure to external security threats, making it ideal for industries such as pharmaceuticals and defense, where regulatory requirements are stringent.
Advantages of Enterprise-Grade AI Tools
Free AI tools typically lack adequate security protocols, and their terms of service allow data to be stored or reused for training purposes. To avoid these vulnerabilities, patent professionals rely on paid, enterprise-grade AI tools that prioritize security and include contractual safeguards tailored to professional needs.
Legal Safeguards to Protect Confidentiality
Strong legal frameworks between the user and the provider of AI-assisted drafting tools do not directly enhance privacy. However, they do provide the user with a strong defense in the event of data theft by demonstrating that all reasonable precautions were taken to avoid negligence. Non-disclosure agreements (NDAs) and privacy clauses can explicitly prohibit AI providers from storing, sharing, or using user data for purposes beyond immediate processing. These measures legally bind providers to confidentiality standards.
Building Trust Through Secure and Responsible AI Use
By combining secure platforms, legal safeguards, and advanced technical measures, patent professionals address security and confidentiality concerns in generative AI. These strategies reduce vulnerabilities and build trust in the technology, enabling its safe integration into workflows. With these measures in place, generative AI can be embraced as a reliable and valuable tool for increasing efficiency and productivity in patent drafting.
Question 5: Do you rather see a democratization of the use of (generative) AI, with AI tools becoming more abundant and cheaper? Or do you think that only big patent law firms will be able to pay for expensive high-end AI tools, giving them a competitive edge over smaller law firms?
Democratization of AI in Patent Drafting
When we talk about a “democratization” of AI-assisted patent drafting, the following aspects come to mind:
- Ease of use: Simple interfaces and structured input and output data formats allow non-experts to use AI without requiring advanced patent drafting skills.
- Affordability: Reducing the cost of AI-assisted patent drafting tools makes them accessible to smaller organizations, startups, and individuals.
- Education and training: Resources, tutorials, and support systems help users effectively understand and use AI-assisted patent drafting tools.
- Open Access: Providing low-cost AI-assisted patent drafting tools encourages widespread adoption.
Many parameters influence the above aspects, and the landscape of AI-assisted patent drafting tools is so diverse that general rules cannot be established. I would instead like to highlight some of the insights I have gained after conducting numerous conferences, workshops, and one-on-one training sessions on using AI-assisted patent drafting software on an international level.
Breaking Down Barriers to AI Accessibility
When it comes to the “ease of use” aspect of democratization, I can say that I have seen inexperienced and fresh patent professionals that benefit greatly from the rigid structure that some AI-assisted patent drafting robots come with, forcing them to provide input data for patent applications in a way that satisfies a set of predetermined rules. For example, in my experience, the best results across the board from all AI-assisted patent drafting tools are obtained when the input data consists of five pieces of data: a set of patent claims, a reference sign list that has only numerals as reference signs, and no different reference numbers referring to an identical term used in the description. Other requirements are drawings that contain the reference numbers in a machine-readable font (and not handwritten), and a list of short descriptions of the drawings, and a reference table that indicates which claims have claim features shown in which figures. Given the task of providing that input data, a patent novice can learn how to draft a decent “beginner” patent application in a matter of days, whereas the same would take weeks or months in the old days of manually drafting patent applications. Having trained numerous patent attorney candidates throughout my career, I can say that it takes much less time to teach these skills than in the old days, and the results are much better.
Balancing Affordability with High-End Capabilities
Regarding the “affordability” and “open access” aspects of democratization, I can say that the most widely used and most powerful AI-assisted patent drafting tool is also the least expensive: ChatGPT. The “Team” plan for ChatGPT provides a secure environment, is priced at $25 USD per user per month, with a minimum of two users, billed annually. Data used under the “Team” plan is excluded from model training by default, giving most users sufficient control over their confidential information. The main drawback of using ChatGPT to draft patent applications is that the user must master not only the art of patent drafting, but also the art of writing efficient prompts for the ChatGPT software. As ChatGPT is only provided as a Software-as-a-Service (SaaS), this AI tool is not suitable for use in all technical fields, and especially not in the fields of pharmaceuticals and defense technology.
Admittedly, if ChatGPT is not the tool of choice for AI-assisted patent drafting, due to data security concerns or due to lack of knowledge and skill on the part of the user, then commercial patent drafting robots are the tool of choice, and they come with significant price tags that disqualify them from being considered “affordable,” thereby limiting access to those who can afford to use them.
The Importance of Education and Training
When it comes to “education and training”, there is a clear gap between a perfect state and today’s reality. There are few resources, tutorials, and support systems available to help users understand and effectively use AI-assisted patent drafting tools. And there is a simple reason for this: the industry of AI-assisted patent drafting tools is developing so rapidly that it does not currently make much sense to provide any sustainable materials for training others in AI-assisted patent drafting.
This is where I come in: I have spent years refining my approach to helping individuals and organizations optimize their workflows using advanced AI tools. Working with nearly fourty (40) different AI robotic patent drafting software manufacturers, I see constant advancements in the field, with new features emerging weekly. This rapid pace has led me to consider offering pre-recorded courses to supplement my live training.
Challenges for AI-Assisted Patent Drafting Training
However, there are too many challenges associated with pre-recorded courses, and I believe that in-person training still has unique value for both training providers and potential attendees. Recorded courses can quickly become outdated as tools and practices change, requiring frequent, resource-intensive updates. They also lack personalization, making it difficult to address individual needs or provide tailored instruction. Learners can struggle with motivation and engagement without real-time interaction, and the lack of hands-on practice often leaves them unprepared to apply their learning to real-world problems. To address these issues, I recommend a blended approach that combines recorded courses with live elements. Regular updates can keep recorded material relevant, while live webinars, Q&A sessions, and forums provide interaction and engagement. The training focus should remain on teaching fundamental AI concepts that are valuable even as technologies evolve. Hands-on exercises and tools can help bridge the gap between theory and application, and personalized support, such as mentorship, can address individual needs.
In summary, while recorded courses are useful for scaling foundational knowledge, they cannot replace the depth and customization of personalized training. A blended learning model can offer the best of both worlds. However, for those who need maximum impact, one-on-one training remains the gold standard, providing customized solutions that align with unique workflows and rapidly evolving AI tools. My goal remains to deliver the most effective training for my clients, whether through in-person sessions, recorded content, or a combination of the two.
The Role of Leadership in Embracing AI
Finally, to return to your original question, “Do you think that only large patent law firms will be able to pay for expensive high-end AI tools, giving them a competitive advantage over smaller firms?”, my answer is straight: no, large patent law firms will not have a competitive advantage over small firms because only they can afford high-end AI tools. From my experience I can tell you that the smaller the firm, the easier it will be to adopt AI-assisted patent drafting, because the larger the organizations, the greater the resistance to adopting AI-assisted patent drafting.
It is not financial power that determines whether or not AI solutions are implemented. It is a mix of elements that must come together – strong leadership, empowered innovators, a supportive culture, and strategic alignment – to effectively adopt and integrate transformative innovations like AI-assisted patent drafting. Take away any one of these elements and the innovation will not be successfully adopted. How do I know this? Because this rule applies to all innovations out there, including AI-assisted patent drafting robots, and I know this from experience because I deal with innovation every day.
Question 6: How do you see the future of the IP professionals, with the impact of (generative) AI?
The Future of IP Professionals with AI
Generative Artificial Intelligence (AI) will transform the future of intellectual property (IP) professionals. While AI will not replace these professionals, it will change their roles, workflows, and required skills in a rapidly evolving field. Its impact will range from automating routine tasks to creating new legal and ethical challenges.
How AI Will Reshape IP Roles
Generative AI has great potential to streamline IP work by automating many labor-intensive and repetitive tasks. For example, AI systems can efficiently draft preliminary versions of patent applications and monitor databases for potential patent and trademark conflicts.
Streamlining Routine Tasks with AI
As AI takes over routine aspects of IP work, the focus of professionals will shift to skills that machines cannot replicate. Creativity, critical thinking, and ethical judgment will become increasingly important. This will also require nuanced human judgment and the ability to interpret and apply laws to novel and unprecedented scenarios.
Shifting Focus to Creativity and Strategy
Rather than replacing IP professionals, AI will serve as a powerful tool to enhance their work. While AI can recommend solutions, draft documents, and analyze data, human oversight is essential to ensure the accuracy and correctness of these outputs. Professionals will play a critical role in refining AI-generated results, creating customized strategies, and providing the nuanced expertise that machines lack because they have no wisdom and will never become wise.
Adapting to a Rapidly Changing IP Landscape
This automation will require professionals to focus less on routine work and more on high-value activities such as interpreting complex legal issues, advising clients, and developing creative solutions to emerging challenges. They will need to provide more strategic, data-driven advice to their clients and improve the speed, depth, and accuracy of their work. Those who are not ready to meet these demands may soon find themselves on the bench.
The Strategic Importance of Human Expertise
In the long term, the role of IP professionals will evolve from being task oriented to being more strategic and consultative. Adaptability will be the key to success. Those who embrace technological advances, build new skills, and focus on areas that require human insight, and creativity will continue to be indispensable in the future of intellectual property.
Question 7: The European Patent Institute has recently adopted guidelines for the use of generative AI. What do European Patent Attorneys, and their clients, need to consider?
European Patent Institute Guidelines for Generative AI
The European Patent Institute’s (EPI) adoption of Guidelines for the Use of Generative AI (click here: https://patentepi.org/en/epi/library/main/538e242d-f1be-46bd-bb34-b948e1544d69/file) is an easy-to-read document that sets standards for European patent attorneys and their work with clients. These EPI Guidelines introduce an overarching principle and twelve (12) rules to ensure that AI is effectively integrated into workflows without compromising legal or professional standards.
The Overarching Principle: Maintaining Professional Standards
The overarching principle, in abbreviated form, reiterates and specifies for the use of AI in professional work that a European Patent Attorney must maintain the highest standards of integrity, ensure confidentiality where required, and put the interests of the client first, as required by Article 1 of the EPI Code of Conduct.
Ensuring Confidentiality and Avoiding Errors
The twelve EPI guidelines, in abbreviated form, are
EPI Guideline 1: European Patent Attorneys should understand generative AI models, focusing on immediate confidentiality and the potential for hallucination in the models they use.
EPI Guideline 2a: European Patent Attorneys should ensure appropriate confidentiality of training data, prompts and other content shared with AI models. If confidentiality is uncertain, the model should not be used.
EPI Guideline 2b: European Patent Attorneys must understand the risks and methods of non-confidential disclosure associated with specific AI models.
EPI Guideline 3a: European Patent Attorneys remain fully responsible for their professional work and cannot use generative AI to excuse errors or omissions.
Aligning AI Use with Client Preferences
EPI Guideline 3b: European Patent Attorneys must review AI-generated work to ensure that it meets the same standards as work produced by a competent human practitioner.
EPI Guideline 4: European Patent Attorneys must determine their clients’ preferences regarding the use of generative AI before using it in their cases.
Transparency in AI Integration and Disclosure
EPI Guideline 5a: European Patent Attorneys may publicly state that their work involves AI tools, provided that such statements are accurate, fair, dignified and non-discriminatory.
EPI Guideline 5b: European Patent Attorneys are not required to disclose the use of generative AI in communications with the EPO or the UPC unless required to do so by binding rules or client instructions. Any disclosure must be accurate and dignified.
Navigating Legal and Ethical Considerations
EPI Guideline 6: European Patent Attorneys shall establish independent user accounts for different clients where confidentiality risks in the AI model warrant this.
EPI Guideline 7a: European Patent Attorneys shall comply with all relevant legislation affecting the use of generative AI, including the European Artificial Intelligence Act 2024.
EPI Guideline 7b: European Patent Attorneys must consider restrictions or obligations imposed by external organizations, such as professional regulators or insurers, that may affect their use of generative AI.
EPI Guideline 8: Fees for work involving AI tools must fairly reflect the time, effort, difficulty and risk, including the costs of setting up, training and reviewing AI-generated work.
I believe the above 12 rules are comprehensive in themselves, and also common sense in light of the overarching principle as mandated by the EPI Code of Conduct.