ChatGPT: An Affordable Solution Or A Genuine Alternative For AI-Assisted Patent Drafting?

Before I begin exploring ChatGPT, let me explain my background.

Sometimes I check if what I thought in the past still holds true. This weekend, I did just that. I rewatched a talk I gave in May 2018 about the future of the patent profession (click here). That was six years ago. I believe this talk is still relevant today. If you're short on time, you can just watch the last 30 minutes. There's a discussion there that brings up some important points. These points are even more relevant now.

I said something in that talk that surprises me now. It was right at the start of the talk:

I repeated the same prediction at the minute mark 03:35 in the talk, which you can see here.

And my prediction turned out to be completely correct. Last year, in May 2023, exactly five years after my presentation at the International Intellectual Property Society (IIPS) in New York, 80% of my work drafting patents was done by a robot. At that time, there were 12 different robots for drafting patents. All of them greatly reduced the amount of work needed for drafting patents.

I have always advised to prepare for future changes in our profession and to invest time and money in one or more technology accelerators. Everyone can invest in different technology accelerators because the field of patents is very diverse and offers many specialization opportunities. The division of labor, made possible by the internet, supports this.

In that talk, I showed a slide that displayed what I had invested my time and money in at that time.

Some of the areas in my above slide went better than other areas but one area made it big time: "drafting figure descriptions for patent applications".

The Year of 2024 Will Be A Boom Year for Automated Patent Drafting

Here is my next prediction: by the end of this year, 50% of all patent applications will be made with the help of generative text software. There are two reasons for this.

The first reason is that the development of Large Language Models (LLMs) is growing exponentially, not linearly or polynomially. In terms of growth, linear growth increases steadily over time, while polynomial growth varies in complexity. However, exponential growth speeds up rapidly based on its current size. I explained this in detail from the [01:08:00] timestamp in my speech at the International Intellectual Property Society (IIPS), which you can see here. Think of it like the water lilies on a lake: the day before the lake is fully covered, it is only half covered.

The second reason is that there are almost free LLMs available for professional use in drafting patent applications. Examples include ChatGPT, as demonstrated by Iain Russel, and LM Studio, as demonstrated by Bastian Best. You can see these demonstrations here and here.

Bastian Best uses the LM Studio software on his laptop while it is completely offline. This solves all confidentiality issues at once.

Iain Russel focuses on the ChatGPT, which is available online for free. This approach is practical because the company behind ChatGPT recently launched a paid "Team" plan that addresses data security and privacy concerns. The "Team" plan costs only 30 US$/month.

Now back to where I started: is ChatGPT an affordable solution and a genuine alternative for AI-assisted patent drafting?

Many are worried about data security.

ChatGPT as a Professional Tool

Check out what OpenAI, the creator of ChatGPT, says about data security on their website: https://openai.com/enterprise-privacy

There are four plans for using ChatGPT: Free, Plus, Team, and Enterprise.

Here are the key points for the Team and Enterprise plans:

• OpenAI does not use your business data to train their language models.
• You keep all rights to the information you provide to their services. You also own any output you get from their services, as allowed by law.
• OpenAI encrypts all data when it is stored (using AES-256) and when it is sent (using TLS 1.2+). They also have strict rules about who can see your data.
• Only certain people at OpenAI can look at your business data. This includes end users within OpenAI, workspace admins who manage workspaces and access, and a few authorized employees. These employees might need to see the data for technical support, to check for misuse of the platform, or to follow the law. Also, some third-party contractors who agree to keep your information secret and safe may review the data, but only to check for misuse and abuse.
• These are strong assurances from OpenAI. I have highlighted the points from the ChatGPT website that stand out in the data security section. There is more information there that convinces me OpenAI has effectively addressed data security issues, especially with the "Team" and "Enterprise" plans of ChatGPT.

For example, OpenAI has a SOC Type 1 certification. This means their system design meets important trust principles. They are also working towards a SOC Type 2 certification. This is similar, but it also checks how well the system works over time, for at least six months. I am confident OpenAI will reach this goal.

However, there is still a risk that OpenAI employees or third-party contractors might share this information. It is important to note that any data entered into ChatGPT under the "Team" or "Enterprise" plan is protected by law as a trade secret, just by the fact that they are OpenAI's employees. The same protection applies to third-party contractors agreeing to keep the information confidential and secure. This is true under "international" law, which means it is the same everywhere.

ChatGPT Input does not Create "Prior Art"

I am completely sure that the statements I mentioned earlier are enough to ensure that any input into ChatGPT under the "Team" plan will not be considered prior art by all the patent offices I work with.

Regarding the European Patent Office (EPO), the Board of Appeal decided in case T 1085/92 that a company's own staff usually cannot be seen as "the public" under Article 54(2) of the European Patent Convention (EPC). This was also supported in cases T 1464/05 and T 1057/09. Other patent offices will likely have the same view as the EPO. I also do not see why a court would think differently.

Therefore, from a patent law perspective, using ChatGPT under the "Team" plan is safe to help write patent applications.

From my surveys among participants in my regular Robot Patent Drafting workshops (click here), 3 out of 10 professional patent drafters openly say they are already using ChatGPT to help write patent applications.

Other Black Hats and Concerns

I am promoting IP services that are rendered over the Internet since 10 years ago, and I believe that I have heard and seen all arguments that may apply. Here is a comprehensive list:

• Data Security: Unauthorized access to or theft of company data
• Data Privacy: Improper handling of personal data
• Money Laundering: Concealing the origins of illegally obtained money
• Insider Trading: Trading a public company's stock or other securities based on material, non-public information about the company
• Fraud: Deliberate deception to secure unfair or unlawful gain, such as falsifying financial records or lying to stakeholders
• Intellectual Property Theft: Stealing or using someone else's intellectual property without permission
• Bribery and Corruption: Offering, giving, receiving, or soliciting something of value as a means of influencing the actions of an individual
• Cyber Attacks: Attacks on a company's IT infrastructure to steal data
• Employee Misconduct: Actions by employees that are outside of business ethics and company policies
• Regulatory Non-compliance: Failing to adhere to laws and regulations

I could fill an entire book with such destructive thoughts. And none holds. Please check out my White Book about the "1st World Robot Patent Drafting Symposium" (click here), a large group of experts has discussed two (2) long days about these concerns and other matters.

Real Data about Robot Patent Drafting Concerns

Just to give you some data points on the internal security standards of the participants in my regular Robot Patent Drafting workshops (click here), all professional patent drafters, here are some survey results:

• about 45% of the attendees do not know what they have to do when they are facing a specific data protection issue in their company, and there is also no information about that in their company's Intranet available
• only about 15% of the companies that the attendees are working for are SOC 1 compliant. The other attendees often did not even know that there exists such a security standard
• only about 30% of the attendees are sending encrypted emails all the time. The other attendees either encrypt only sometimes or not at all.
• 85% of the attendees use cloud storage, and not proprietary on-site solutions for storing their business data, among them Microsoft, Amazon Web Services (AWS), Google, and Dropbox, in that order of frequency of use

It is unrealistic to expect a higher degree of security from a Software-as-a-Service (SaaS) patent drafting robot service than one applies to other SaaS services.

Special Case: Data Theft in Patent Matters

The concern of data theft during the drafting or filing of patent applications, while valid, can be seen as unrealistic due to several factors. Firstly, it is important to note that patent data is set to be published 18 months after the filing date. This public disclosure is a fundamental aspect of the patent system, designed to share knowledge while providing legal protection to the inventor. The eventual publication of this information diminishes the value of stealing it, as the details of the invention will become publicly accessible in a relatively short period.

Secondly, once a patent application is filed, it gains legal protection. This means that even if the data is stolen after filing, the original filer has legal rights to the invention. The act of filing the application establishes a priority date and a legal record of the invention, making it challenging for a third party to claim ownership or file a patent for the same invention without facing legal consequences.

Lastly, the most vulnerable period for data theft is the time between generating the invention disclosure and filing it with the patent office. However, the practical utility of stealing such information is questionable. If competitors were to acquire this data, using it to develop their own products would involve significant risks. Implementing a stolen invention not only requires substantial time and resources but also poses a high risk of legal challenges once the original patent application is published. Additionally, any product developed using the stolen information would be infringing on the patent rights once the application is granted, leading to potential legal disputes and damages.

Yes, one may be able to conceive a situation where data theft is possible during the patent application process, but the relevancy and likelihood of such an occurrence are quite low. As a patent attorney with 28 years of experience, I have not encountered a single case where this has been a significant issue. This observation underscores the point that while the theoretical risk of data theft exists, its practical impact in the field of patent law is minimal. The combination of eventual public disclosure, legal protections post-filing, and the practical challenges of illicitly using stolen information makes such an event not only unlikely but also of questionable benefit to the perpetrator. Therefore, concerns about data theft in the context of drafting or filing patent applications, while understandable, do not align with the realities and experiences within the field.

In summary, the risk of data theft in the context of patent applications, while not entirely dismissible, is mitigated by the inevitable public disclosure of patent information, the legal protection afforded by filing an application, and the practical challenges and risks associated with illicitly using stolen invention details.

And If there is an Espionage Threat?

When there is a real worry about spying, the situation is different. Spying involves careful and complex ways to get secret information. These cases are rare but can threaten the safety of important patent information. In your work, you might come across such a case. These risks are not common, but they are real.

Keeping digital information fully safe is a big challenge. Even with strong cybersecurity and using on-site servers and on-site cloud computing, digital systems have weaknesses. Skilled people or groups who are spying can use these weaknesses. In these cases, using only paper documents and not using computers can be a wise choice.

Using only paper documents and avoiding digital tools makes it harder for someone to attack through the internet. This old-fashioned method in our digital world offers better security than digital systems, especially against targeted spying. But, it is important to understand that this method has problems, too. For example, managing documents, working together, and doing things efficiently is harder.

In summary, it is hard to ensure that digital systems are completely safe from spying in patent law. Using paper documents can be a good alternative. This method has disadvantages, too. It may not be necessary or practical for most patients. Still, it can be useful in very sensitive cases with a big risk of spying, such as in military techniques or early medical drug development.

Conclusion

This is the third reason why I believe my prediction will come true. By the end of this year, 50% of all patent applications will be made with the help of generative text software. There is a need for automated help in writing patent applications, and a solution is readily available: ChatGPT. It is affordable and safe enough.

So, to answer the question in the title of this article: yes, ChatGPT can be seen as a budget-friendly tool for drafting patents, but it is a not very effective one. It is just better than not having any AI-assisted tool for drafting patents at all.

Will there still be a need for specialized Patent Drafting Robots? Yes. ChatGPT is quite good, but using a professional tool makes a difference. I say this as someone who has been a patent attorney for 28 years and still drafts patent applications professionally, but today with the help of AI-assisted patent drafting tools.

And if you are into patent drafting yourself, I recommend that you start with what is listed on my website (click here). Sign up for one of my upcoming workshops (click here). Or sign up for a 1-to-1 training with one of our trainers (click here).

Don't wait, do it now.

Martin "AI" Schweiger