Letters of Intent

Pankaj Raval (00:03)
Hello and welcome back to Letters of Intent. This is your podcast for deal makers and risk takers. So if you're one of those people, you're in the right place. I am your host, Pankaj Raval, and I'm joined today by my co host and trusted advisor here, Sahil Chaudry. Sahil, how are you doing today?

Sahil Chaudry (00:19)
I'm doing great.

Absolutely. I'm excited to tackle today's episode. We're talking about giving our community an AI survival guide. And this is a really hot topic. I just went to the press published conference, the Hollywood Creators Summit in LA. And it was a major topic of discussion, especially by the YouTube team. So the event was one of the main sponsors of the event was YouTube and Gemini, and they had a YouTube diner where you could

Pankaj Raval (00:29)
Yeah.

Sahil Chaudry (00:46)
Go and talk to the YouTube team about a creator. And one of the most important tools they gave me was the YouTube Creator Studio. it's incredible. the YouTube Creator Studio is designed to help creators come up with concepts, analyze the data, analyze their viewership, but it gives you scripts, it gives you thumbnails, it gives you titles. So it's a very powerful tool, but it does beg the question.

Where does your ownership begin and where does AI's authorship begin?

Pankaj Raval (01:16)
Absolutely. It's a big question. It's a very big question that I think we've discussed in the past is not settled. There's gonna be a lot of litigation, there's gonna be a lot of discussion. we'll see how much actually regulation there's gonna be. This administration's not really interested in regulating this area, but it's gonna need regulation, right? Because it is really very much the wild, wild west right now. but I think there are some

important learnings we've seen even over the last few years in this kind of nascent and world of AI and leveraging AI chatbots, which has really taken over the internet well it has really changed the way we operate, like very few other things there.

Sahil Chaudry (01:48)
Well

The new development is that human beings are not the only ones capable of authorship. You can generate code, you can generate copy designs, which you were able to seek copyright protection for your original ideas. Now you're putting your ideas into an AI platform which is using your ideas as part of its training So you could be unintentionally using other people's

Pankaj Raval (01:55)
Right. Right.

Sahil Chaudry (02:14)
And not only that, what you are generating could be based on other people's content. So the question we want to answer or at least help our community answer is how do you determine where your chain of title begins and ends? And it's not a satisfying answer because the law is unsettled here, but we can give our community here

Some best practices. For example, when we're talking about the ownership gap, since we know that courts have held human authorship has a collaborative element with AI, and the human elements are copyrightable, you need to make sure you're documenting.

The ways that you as a human being are interacting with the AI, what are the original ideas that are going into your prompts? So, for example, a very practical step here to evidence your chain of title is to save your prompt history and make sure that you have some evidence of the original ideas you've put into AI.

The second thing that we would suggest is make sure you're using an enterprise level software. They can be expensive, but that's the reality that similar to let's say a business associate agreement when you're dealing with HIPAA compliance in the medical field, the AI platforms at the enterprise level have contractual language that protects your confidentiality.

Usually. And you do need to investigate which platform you're using and how they're protecting your data. But generally the enterprise levels of these software platforms are protecting confidential information. So I know Pankaj you deal a lot with chain of title when it comes to IP. How are you seeing clients deal with this question and what would you recommend?

Pankaj Raval (03:53)
Yeah, the truth is don't think clients are thinking about it enough. Most clients aren't. I would say ten percent do ask the question, but ninety percent are just going full steam forward. which I think is one strategy, I don't know if it's always the best strategy because the issue with all these legal anything legal is always it comes up after the fact. It doesn't happen immediately, right? No one hears immediately, but later on when

contractors question when your IP is questioned, when, in litigation, worst case scenario, you're spending, tens, hundreds of thousands of dollars on litigation. And all of a sudden now you find out you have a vulnerability because you used AI in a way you didn't realize could compromise your IP. So I think we've gone over a little bit in the past, but I think it's a worthwhile refresher about like just the universe of IP specifically. And then we could talk about other areas of

How AI is being used. But if you're a business owner, IP is a very important area that you need to understand and focus on. And when you think talk about IP, there's four main areas of IP. You're talking about copyrights, you're talking about trademarks, you're talking about patents, and you talk about trade secrets. Okay. and within those, there's a whole universe, right? People generally focus on one area. Like example, I focus mainly on trademarks, we do copyright as well. But the thing is like they're all very complex areas of law.

and a lot of nuance there. So when you talk about copyright, we're talking about okay, copyright is the tangible expression of a idea. Okay. It's not the idea itself, but it's the tangible expression of idea. But one of the requirements of copyright is that the work be generated by a human. Okay. but then it gets complicated because like okay well what if a human prompted the AI, the AI created it. How is that different from using Photoshop?

Right. And the courts have looked at this, right? And courts are looking at this and they're probably gonna say, yes, it is somewhat different because a prompt versus you using a pen to draw something on a digital tablet is different. But then now digital tablets are helping you, draw a smoother line, draw a more, refined image, right? So these are very fact specific, nuanced questions that will need to be addressed in the future.

But for business owners, you need to realize that if your work product is primarily AI generated and doesn't really have much human and this is something where you may need to also like look at the record, right? Like when you're submitting these, you need to show the copyright office how much human effort went into this, right? Maybe you spent hours prompting this to get to where you want it to go. Maybe, that's sufficient. But I think it will

of a multi factor analysis that the copyright office and courts are using. Then talk about trademarks. Yeah, go ahead.

Sahil Chaudry (06:12)
So

no, so outside of saving your prompt history, what are some other ways our clients can prove their authorship?

Pankaj Raval (06:20)
Well, taking some action before filing a copyright of the hand drawing, right? Maybe you draw it out a crude hand drawing and you copyright that and then you prompt it because then you would own all derivative uses of that image. So it could be, anything hand drawn and then you could file a copyright for that. There's ways you could like mail yourself that. There's kind of the poor man's copyright or patent where

you create something, you mail it to yourself, certified mail. So there's like a time stamped documented copy that's verified by the post office that says, yeah, I created on this date and send it to myself and this is it. And maybe that'll help you prove that you created it on those dates and then any derivative uses of that would still be yours.

Always best to file, but it's not always practical to file every copyright, every image, everything you've ever created because that can be expensive and time consuming. So those are ways to think about it.

Sahil Chaudry (07:06)
let's say I'm an employee and I'm working for a biotech company and I'm using Chat GPT, but part of my prompts include the company's trade secrets. Is there any risk of public disclosure there, could kill my patent or trade secret protections?

Pankaj Raval (07:21)
Absolutely. Absolutely. Yeah. What companies wanna do is have prompting guidelines and also have a closed system. At the end the day, you need, especially enterprise level, you need to have a clearly closed system where there's maybe an NDA in place with a provider. it's clear that anything you guys use this for is covered by this NDA, but also that there's no

any kind of personal identifying information or client identifying information is redacted. There are services out there that lay on top of AI platforms that will help you redact called PII, like, patient information or other confidential information. in healthcare you really gotta be aware of this. But in many different sectors you gotta be aware of what kind of confidential information is being released out there. And making sure that, you're working

with a large provider that will ensure that your information is kept safely. But the reality is that, we've been using cloud services for the last more than decade, right? Like Google, Drive, OneDrive, Dropbox, all these cloud services that are susceptible to the same potential disclosure, right? Right now, those are your private documents, but the question is, could they get out there? What if there's a hack? So there's always that risk.

So you wanna make sure that also to the extent possible, you're limiting how much confidential information is even being released on these platforms.

Right. So fundamentally, and I'm not a computer scientist, I can only tell you what I understand from a lay person's approach, is that they do not share that information for training. They don't use that information that you're uploading for training. That's what they're saying, okay? and we have to take them at the word. I don't know if anyone's actually looked into that deeper. I'm sure larger enterprises maybe have.

But they don't use that information for training and any information that potentially is also encrypted. So they should be using high-level encryption so that they cannot see what you've been uploading. So at the end of the day, they can't tell what it is that you've been uploading. It's only personal and proprietary to you and your workspace. So that's what they should be using, that's what they should be implementing. who's verified that, who's checked that?

That's another question to solve.

And then I think part of that too, is like make sure you have if the options there, opted out from training. Right. So that's another important element you could do. Even if you're using the free version of I think Claude or Chat GPT, you can opt out of training. So that's usually in a settings.

So I think these are all, very important concepts and issues to remember when you're thinking about using AI in your business. Anything from looking at IP to just how you interact with these platforms is important. And also people should know that, hey, if you're also asking questions about any legal

You want to be careful there too, because you want to make sure that's not discoverable. So what you ask, there's been issues where clients asking, divulging information to the AI cases has become subject to discovery. So just be careful about like, what you put in, how you ask these questions. Not to scare anyone and say, hey, don't do it, because these AI tools are very useful and very powerful and especially helping understand issues.

But you need to just be aware of how information's being used and where it's going. And just have a certain amount of everything a little bit because we still don't know exactly what's all being used and how, and it's important to ask these questions.

And I think, everyone, it's still kind of early in the game for AI. so many changes coming. We just saw it I think it was yesterday Nvidia and Microsoft are teaming up for a new AI laptop where you can now run a lot of these right on your own computer. So there might be like a private LLMs in the future. So, it's still an evolving landscape. but it's important to

legal implications at every step of the way, just so to make sure you're protected and don't step into any kind of legal issues that want to deal with. can continue to focus on your business and growing your business.

Thank you again, Sahil. Great episode. Hopefully everyone found this helpful. And like, share, follow for more. We always love to hear from our listeners too. There's anything you wanna hear from us. And until next time, wishing you the best with your deals and taking smart risks.