Straight to the Prompt: IP Lawyers Must Develop AI Skills NOW

“Sheikh’s experience already gives us an example of ChatGPT ‘beating’ humans today in an area we normally consider the core value of human lawyers: client care…. AI is always ‘on.’ It is never tired, frustrated, or egotistical.”

AI promptIn September 2023, one man grabbed the authors’ attention with his astonishing story about defending his trademark registration from an opposition by professional trademark attorneys using ChatGPT. His months-long battle began in December 2022, less than a month after the public launch of the now infamous AI chatbot.

Nine months later, Jamiel Sheikh — an entrepreneur, tech-guru, and adjunct professor — survived the pressure from formal proceedings and obtained a settlement from his opposer without spending a dime. As young trademark attorneys, we were horrified yet extremely curious about what he had done. This article is the result of speaking with Sheikh about his experience and the evolving needs and expectations of sophisticated legal service consumers.

Background – Applying for INSTAMINT

Sheikh is the CEO of Scifn, which specializes in delivering courses, holding events, producing media, and building products at the intersection of AI, blockchain, and finance. He is also an adjunct professor teaching graduate-level machine learning at Columbia Business School, NYU, and Zickin Business School.

In 2021, Sheikh started Instamint, an “enterprise digital assets management platform.” According to Instamint’s website, Instamint allows enterprises to “rapidly and economically mint and manage digital assets,” like invoices, shipping containers, and smart contracts. That November, Sheikh applied to register the trademark INSTAMINT for uses related to financial services on the blockchain. Sheikh’s usual practice as a serial entrepreneur was to file for trademarks by himself, get them awarded, and then add an attorney’s name to the registration.

That is not what he did this time.

Facebook’s Opposition and ChatGPT’s Defense

A year later, in November 2022, the U.S. Patent and Trademark Office (USPTO) notified Sheikh that the trademark application for INSTAMINT was approved and would be published and available for opposition for 30 days. Not long after, Sheikh received notice that Facebook had requested an extension of time to oppose. After a few extensions, Facebook filed its notice of opposition in April 2023.

Instead of asking his trademark attorney to handle the opposition, however, he decided to represent himself, armed with OpenAI’s newly released generative AI tool. Sheikh explains his rationale as follows: “I decided that I could spend $50,000 and end up losing this lawsuit, or I could choose this, as something I can learn from—how AI tools, specifically ChatGPT4, could be incorporated into a real-world situation.”

That said, he had invested an intangible amount of blood, sweat, and tears into developing brand recognition for the INSTAMINT mark in his target community. Defending his trademark registration by himself was not a simple cost-benefit decision, nor was it just an experiment to him. In fact, Sheikh explained that he originally looked for an attorney to represent him. In looking for the right attorney, he interviewed several candidates but none seemed to understand what his product was and none seemed like they would truly fight for his “baby” like he would.

So Sheikh’s defense began. The first step was, of course, to answer the opposition. He relied on ChatGPT for the basic rules and for drafting some language. To lawyers, the result was highly unorthodox, to say the least. In addition to the obvious formatting issues related to the margins and alignment of text in Word (which were due to his Microsoft Word skills, rather than ChatGPT), his responsive paragraphs did not begin with the standard “admitted” or “denied” responses. In fact, his paragraphs appeared not to correspond one-to-one to the Opposer’s paragraphs.

For example, in response to the (boilerplate) allegation “15. Applicant’s Mark has caused or is likely to cause dilution of the distinctive quality of the INSTAGRAM Mark . . . ,” he answered:

  1. The prefix “Insta” is a common component used by various companies across different industries and should not be granted extensive protection. The widespread use of the “Insta” prefix indicates that it carries minimal distinctiveness and is unlikely to cause confusion when used in conjunction with other elements, as suggested in King Candy Co. v. Eunice King’s Kitchen, Inc. 496 F.2d 1400 (C.C.P.A. 1974).”

In case you were wondering, King Candy is not a hallucination; it is an actual case that has roughly 4,000 citing references.

Sheikh did not comply with the Trademark Rules of Practice and Federal Rules of Civil Procedure. The Trademark Trial and Appeal Board (TTAB) typically requires strict compliance with the rules, whether or not they are represented by counsel, so his response to Facebook was likely deficient. But the TTAB accepted Sheikh’s pleading this time. The number of paragraphs did match and the responses can arguably be construed as denying each allegation, albeit in an unconventional and argumentative manner.

Having survived the first hurdle, Sheikh continued consulting ChatGPT, not only for format, but also for background information and general strategic advice. He gave ChatGPT several comprehensive prompts, such as “Party A has registered mark x for a specific use of technology services under class 35 within the context of social media. Party B has registered the same mark under class 35 for a different use. Party A now applies to extend the uses of the mark for t-shirts and beverages. If registered, does this preclude Party B from creating t-shirts with its own brand on it? As opposed to being a t-shirt manufacturer?”

He even consulted ChatGPT on the tone he should strike in informal email communications with opposing counsel.

“I created this massive prompt with lots of conditional logic that was well-structured. It forced ChatGPT to come up with strategies and explain to me the rationale for those strategies. Writing the appropriate prompts are critical to implementing ChatGPT effectively. If they’re too complex, they could fall apart, but if they’re the perfect length and detail of comprehension, they could be effective and accurate.”

On his social media, Sheikh shared 20–30 prompts (out of hundreds) that he used.

ChatGPT did more than regurgitate rules or summarize case law. According to Sheikh, it helped him appreciate some of the “meta”-aspects of the legal practice: “I came to understand what Facebook’s trademark strategy is, how Facebook works its trademarks, analyzing their current and old filings and understanding where they like to negotiate and where they may not, and the intricacies of the U.S. trademark process.”

By September 2023, Sheikh had exchanged initial disclosures, filed for extensions in other Meta trademark applications, and went back and forth numerous times by email with opposing counsel, all while consulting ChatGPT.

Sheikh ultimately accepted Facebook’s offer to drop the opposition in exchange for amending the use of his INSTAMINT mark to exclude social media—an area he was not engaged in. He filed the motion to amend with ChatGPT’s help and the USPTO dismissed the opposition on September 27.

Lessons Learned

Let’s get the obvious and clichéd takeaways out of the way first. As of 2023, ChatGPT (versions 3.5 and 4) cannot replace a dedicated trademark attorney. There are significant risks of inaccurate advice or hallucinations, not to mention, “doing it yourself” remains extremely time-consuming and stressful for both pro se clients and seasoned in-house counsel.

“While ChatGPT was there to help, the emotional toll of having to go through such a process on my own was steep,” said Sheikh. “There were times I wanted to cancel my trademark and just walk away. Every time I received an email from Facebook’s lawyers, my anxiety level would rocket.”

Also beware, depending on the terms of service, communicating with AI may breach confidentiality obligations and inadvertently waive privilege.

But Sheikh’s experience already gives us an example of ChatGPT “beating” humans today in an area we normally consider the core value of human lawyers: client care.

“I think we are crossing that boundary in the world. There is a comfort zone with ChatGPT; it communicates with me and educates me in a way I’m comfortable with wherein there is an absorption of information maximized and translated back to me, and every time you submit a prompt, an entire history is sent around that prompt,” Sheikh said. “Every time I put more knowledge into that model, the model is updated and caters to my communication with it. It’s an unprecedented technological time in our history.”

Like autonomous vehicles, AI is always “on.” It is never tired, frustrated, or egotistical. It has perfect recall, it is not necessarily bound by conventional thinking or biases, and it offers consistent quality service. It is not altruistic, but its motivations are more transparent and rational than a human’s.

So if the risks of hallucinations are eliminated in the future, how are lawyers supposed to provide competitive value in the age of AI?

First of all, you should begin learning AI as a necessary tool for your practice and do it fast.

If you have dabbled with ChatGPT and concluded, there’s nothing to worry about, it’s possible you haven’t given AI enough of a chance (it’s also possible you’re using free, second-rate technology). The answers from ChatGPT are only as good as the questions you ask. Asking ChatGPT basic questions like “how to prosecute a trademark” or “how to respond to an office action” won’t be more helpful than a simple guide on the USPTO website.

If your opposing counsel isn’t using AI now, they will in very short order. Students of law, not to mention grade school students, are growing up prompting ChatGPT for solutions, nay ideas, to complex problems. As important, clients are increasingly relying on ChatGPT and other AI in their daily operations (and in their interactions with their lawyers).

As Sheikh puts it, “I would hire a lawyer next time, but at the same time, I would also expect my lawyer to have a similar level of sophistication of the AI programs that efficiently represent me and my business, and for my lawyer to have a high level of prompt engineering on how to best use the latest AI programs.”

How should you start to develop prompt engineering skills? Like all aspects of law practice, trial and error is the best advice. Find and invest in the specific AI program that fits your area of work, and then generate prompts as often as possible, tinkering with them, to achieve the best and most accurate results like building your repertoire of high-use forms. And read articles about how to better draft prompts, just as you would read Bryan Garner or Point Made, and learn best practices from the elite.

Second, stay up to date.

As these technologies continue to evolve, we need to ensure we keep up with the latest versions.

According to Ray Kurzweil’s “law of accelerating returns,” the next advance in technology will always be bigger than the last (or put differently, the next advance will always come quicker than the last).

For example, ChatGPT 3.5 and 4 were released in late 2022 and early 2023. Rumors are now swirling about the release of ChatGPT 5 soon. AI grows exponentially fast because existing AI is used to train the next AI. Sheikh hasn’t used ChatGPT 3.5 in months. ChatGPT 4 will also be ancient technology soon. We need to keep up with the latest versions because we know our clients are using the latest technologies as well.

Richard Susskind predicts that, “by 2040, to pick a date two decades or so hence that will be mid-career for young lawyers of today, it is neither hyperbolic nor fanciful to expect that the legal profession will have changed beyond recognition.”

Finally, the key ingredient to providing tangible value is connecting with your clients.

At least for now, AI is generating text based on our collective knowledge in the digital universe. AI can provide you information and complex data analysis, but it doesn’t have authentic lived experiences. One day, it may provide perfect text, audio, and video answers (prompted or unprompted), but it cannot relate to the messy real-world experiences of human enterprise. Only a talented, worldly, and empathetic human attorney can.

So don’t forget to approach your human clients as partners in this extremely challenging environment. Acknowledge the reality of AI and build a sustainable business relationship with your clients together.

Image source: Deposit Photos
Author: zentro
Image ID: 308474558 


Warning & Disclaimer: The pages, articles and comments on do not constitute legal advice, nor do they create any attorney-client relationship. The articles published express the personal opinion and views of the author as of the time of publication and should not be attributed to the author’s employer, clients or the sponsors of

Join the Discussion

2 comments so far.

  • [Avatar for Michael A. Cicero]
    Michael A. Cicero
    December 6, 2023 12:41 pm

    Messrs. Yoshikawa and Zhong – your article was very interesting. Thank you for sharing your insights here.

  • [Avatar for OptiIP]
    December 6, 2023 11:56 am

    >>>Sheikh ultimately accepted Facebook’s offer to drop the opposition in exchange for amending the use of his INSTAMINT mark to exclude social media—an area he was not engaged in.

    I get a feeling it shouldn’t have taken a glorified text-predict bot, let alone a $50,000 lawyer, to have gotten him to that point. If he had no intention to get into social media, he could have easily narrowed his initial application out of that field. *Everyone* knows what Instagram is.