Despite the risks, AI can be a useful partner in patent prosecution. Shwetangi Sinha and Vasu Bral of Sagacious IP offer a useful guide for patent lawyers.
The emergence of large-scale language models (LLMs) and advanced generative AI, such as OpenAI’s ChatGPT (“Chat Generative Pre-trained Transformer”) and Google’s BERT (“Bidirectional Encoder Representations from Transformers”), has generated significant interest due to their ability to generate human-like responses across diverse topics.
These LLMs have great potential to revolutionize various industries, including IP.
In the intellectual property field, generative AI technology has great potential to streamline a variety of tasks, greatly improving the efficiency and accuracy of these tasks, including prior art analysis, trademark searches, and automated contract drafting. However, where this technology really shines is in patent drafting.
Generative AI models such as ChatGPT have made impressive progress in generating human-like text, and as these models continue to evolve, their integration into the IP field creates an opportunity to automate the drafting process while still providing due diligence.
GenAI's Role in Patent Drafting
An effective patent draft accurately captures the novelty, originality, and usefulness of an invention while ensuring compliance with patent law. A properly drafted patent protects an inventor's intellectual property and serves as a valuable asset for licensing, commercialization, and enforcement.
ChatGPT uses its reasoning capabilities to assist in this complex process, translating an inventor's concept into a structured application that maximizes the scope of protection, offering several advantages including rapid drafting, integration with human expertise and identification of alternative implementations.
Although some AI enthusiasts claim that AI tools can write complete patent applications on their own, this expectation is often unrealistic. The patent application writing process begins with extracting actionable information, which requires a comprehensive technical understanding of the invention under consideration. This step is especially important for complex documents involving cutting-edge technologies, where ChatGPT may face challenges.
Still, AI can bring valuable contributions.
Best Use Cases for AI in Patent Drafting
While AI cannot be relied on alone to prepare a final patent application, it can certainly provide a valuable advantage. AI is clearly good at understanding and manipulating large amounts of text. In the patent preparation process, a variety of data is typically available, including invention disclosures and references to prior art (patents, research papers, etc.). Thus, AI can help:
- Comparison of the invention with the prior art: ChatGPT helps to conduct a comparative analysis of inventions and prior art, highlighting novel features and clarifying the scope of patents. This feature streamlines the early drafting stages and provides valuable insights for patent lawyers and drafting professionals.
- Generate the first claim: AI can provide a foundational framework for a patent application and generate a concise initial set of claims to guide subsequent drafting. These initial claims can then serve as a starting point for human refinement and expansion, meaning patent attorneys and drafting professionals may not need to start from scratch every time they begin work.
- Supplemental Research: The LLM can serve as a valuable tool for supplemental research, similar to how patent attorneys currently use resources such as web searches, Wikipedia, and academic articles, but its use should be limited to text generation for non-innovative sections of an application, such as boilerplate sections, state-of-the-art descriptions, and contextual information.
Observation: Asking the AI to generate a comprehensive set of claims may result in a loss of quality, but smaller claim sets are more efficient.
Possible reason: This is likely because the AI response time remains the same regardless of how many claims you need to generate. Whether you're creating one claim or multiple claims, the AI takes the same amount of time for each.
Therefore, as the volume of claims increases, the AI may not have enough time to deeply investigate each claim, leading to less comprehensive claims and lower quality.
Addressing these considerations and effectively leveraging AI capabilities can significantly improve the efficiency and accuracy of the patent drafting process. Collaboration between AI developers, legal experts, and innovators is essential to responsibly advancing the role of AI in intellectual property management.
Potential pitfalls and considerations
Despite their capabilities, it is important to be aware of potential pitfalls and considerations when integrating generative AI into your patent drafting process.
- compliance: LLMs may lack a nuanced understanding of legal complexities, which can lead to errors and omissions in patent applications. Legal professionals must closely review and validate AI-generated content to ensure compliance with patent laws and regulations and mitigate legal risks.
- Limited contextual understanding: Although LLMs are good at pattern recognition, they may lack domain-specific contextual understanding and may miss technical details and the broader context of an invention.
- Accuracy Challenges: Ensuring accuracy of AI-generated content can be a challenge in maintaining the integrity of a patent application: one inaccuracy can lead to a cascade of errors, making it difficult for inventors to thoroughly review all documents, especially given the presence of legalese and patent-specific terminology.
- Suboptimal target training: Despite attempts to target training AI models using patent documents, the results are often suboptimal, highlighting the complexities of patent drafting that require human expertise.
Therefore, excessive reliance on AI without human intervention can undermine the overall quality and completeness of patent applications. Striking a balance between leveraging AI for efficiency and maintaining human oversight is essential to ensure accuracy, novelty, and comprehensiveness of applications.
Security Considerations
When dealing with new inventions and proprietary information, security and data privacy are of paramount importance, and recent data leaks from public AI tools such as ChatGPT have understandably raised concerns.
Most LLMs are run by private companies and hosted remotely as cloud-based services. The terms and conditions of end-user agreements with third parties will almost certainly state that the third parties will collect and store your conversations with the LLM. Additionally, the stored data may be used to train future iterations of the LLM.
This means that if you hire a Master of Laws to draft the claims or specification for any aspect of your claimed invention, you may be disclosing this invention to third parties, which will not only trigger the 12 month grace period for filing in the U.S., but may also prevent you from filing in countries or regions that require absolute novelty.
To mitigate these risks, we recommend not inputting sensitive invention disclosures or proprietary data into publicly available AI tools, and instead consider leveraging Large Scale Language Models (LLMs) such as GPT-4, Claude, or PaLM (Pathways Language Model) via secure API access.
- Sensitive API Access: All major AI providers, including OpenAI, Anthropic, and Google, offer confidential API access to their LLM models. These confidential APIs ensure that input data remains secure and is not used to retrain models or for other unintended purposes.
- An open source model for private data centers: Another approach is to use an open-source LLM model hosted in a private data center or cloud provider such as AWS or Google Cloud. In this setup, the LLM model and your proprietary data never leave a controlled environment, greatly reducing potential risks.
Patent professionals must carefully evaluate their security needs and comfort levels when integrating AI into their workflows. Implementing robust data protection measures and carefully balancing risks and benefits is essential to responsibly harnessing the power of AI for patent drafting and IP management.
What ChatGPT can and can't do
Features and limitations
- Rapid Draft Generation: AI expedites the drafting process, producing patent applications quickly and efficiently.
- Limited understanding of new technology: AI may struggle with recent technological advances.
- Providing an outline of the claims: AI provides a structured outline, helping to organize and make your application consistent.
- Inaccurate description without human supervision: AI-generated explanations may lack accuracy without human oversight.
- Identifying alternative implementations: AI suggests alternative implementations and encourages creativity and exploration.
- Lack of originality or novelty in the invention: Because AI relies on historical data, it may not be able to generate truly new solutions.
- Assistance with non-innovative parts of the application: AI handles the non-innovative sections, freeing up time for humans to focus on the innovative aspects.
- Inability to fully address legal complexities: AI may not be able to fully understand the complex legal nuances of patent applications.
- Generating background, summary and benefits text: AI generates text for different sections, providing context and highlighting the benefits of your patent application.
- Potential security risks regarding confidential information: AI poses security risks when dealing with sensitive information in patent applications.
Final thoughts
While generative AI can bring significant benefits to patent drafting, including speeding up the process and providing valuable assistance in tasks such as drafting patent claim outlines and identifying alternative implementations, it also has limitations.
These include the potential inaccuracies of AI-generated explanations, the difficulty of understanding complex legal nuances, and the risks of over-reliance on AI without human input.
Despite these challenges, AI remains a valuable tool in patent drafting, and when used in conjunction with human expertise, it can increase efficiency and productivity. To protect sensitive information, it is important to address security concerns associated with AI tools.
Ultimately, the integration of AI into patent drafting must be approached carefully, balancing its capabilities and limitations to optimize efficiency and accuracy.
Shwetangi Sinha is the Project Manager for ICT Drafting and Litigation. Wise IP
Vasu Bral is the Project Manager for ICT Drafting and Litigation. Wise IP
