
The Hidden Legal Risks of Using AI with Peter Nieves | Part 2 #973
In Part 2 of the Norris Group Real Estate Podcast, Joey Romero continues his conversation with Peter Nieves, an intellectual property and patent attorney, about how real estate investors and entrepreneurs can take advantage of AI without putting their businesses, confidential information, and intellectual property at risk.
Peter explains why using the right AI tools matters, especially when handling sensitive client information, financial records, proprietary data, and business strategies. He discusses the importance of local AI systems, protecting trade secrets, understanding software ownership, and avoiding copyright infringement when using AI-generated content.
The conversation also highlights why AI should never operate without human supervision. From inaccurate information and AI hallucinations to third-party developers claiming ownership of software, Peter shares practical considerations businesses should understand before investing in AI technology.
He also shares his prediction for 2027: the Copyright Office may recognize certain AI-generated works as eligible for copyright protection when sufficient human creativity is involved.
In this episode:
How to protect confidential business information and trade secrets when using AI.
Why local AI systems can offer better data protection.
Copyright infringement risks when using AI-generated content.
Why AI-generated information must always be verified.
What to consider when hiring third-party AI developers and establishing software ownership.
The importance of creating an AI business plan with clear usage policies and safeguards.
Peter’s prediction about AI-generated content and copyright protection by 2027. How to use AI responsibly while protecting intellectual property and business assets.
Episode:
Narrator
Welcome to the Norris Group real estate podcast, a show committed to bringing you insights from thought leaders shaping the real estate industry. In each episode, we'll dive into conversations with industry experts and local insiders, all aimed at helping you thrive in an ever changing real estate market, continuing the legacy that Bruce Norris created, sharing valuable knowledge and empowering you on your real estate journey. Whether you're a seasoned pro or a newcomer, this is your go-to source for insider tips, market trends, and success strategy.
All right, welcome back to the Norris Group real estate podcast for part two of our interview with Peter Nieves, intellectual property and patent attorney. Let's get to it. So let's bring it back to our audience, which is small real estate investors.
So let's say somebody uses AI to get either a proprietary seller's list or an underwriting, let's say they create an underwriting model. What's one thing they can do to protect those things that they create? Well, the first thing they should do is they should be used, if they're going to use AI, they have to make sure they're using a confidential version, right? So like use local AI or use enterprise level, a high level of enterprise level, right? So that your data does not go to third parties. I don't even like putting trade secrets on enterprise, I'll be honest, because even though it's not training the model, the information's going to third party, which can result in your loss of trade secrets.
So what I would do instead is I would get local AI, which is an AI model literally being on your local PC that doesn't transmit anything to a third party. It doesn't go to any server, it only stays on your local PC. Nobody's talking about that, by the way, that's a great resource.
You can train the model with your own information, you can get smarter and smarter, and it's not transmitting to any third party, so you don't really run into a lot of legal problems. And then you can put your customer lists in there, you can put their data, their everything from their financial records to healthcare to legal, you do whatever you want, because it's all local, like it's not going anywhere. So the first thing I would do, again, is if I'm using AI, select a platform that is going to maintain the confidentiality of my client list, because client lists fall into the category of trade secrets, right? But you need to be able to prove that you're maintaining it as a trade secret.
So you have to take the steps to maintain the confidentiality, and if you're throwing it into OpenAI's chat GPT, you lost your trade secret. I mean, look at Samsung, Samsung did this, they had two employees who uploaded, two engineers that uploaded source code, trade secret source code of Samsung into OpenAI, and gone. It's immediately like on message boards, and like the whole thing's on GitHub, like right away, right? It's just, there's nothing you can do, it's, and you can't just delete it, like, because it's, again, it's a neural network, it's kind of like taking, if you're making sauce, and you put garlic cloves in the sauce, right, even if they're whole cloves, and you and you cook the sauce, and then you find out one of the people that's invited to the party is allergic to garlic.
I mean, are you going to go and just grab the garlic cloves out and go, okay, it's safe? No, because the oils of the garlic already in the sauce, like you can't remove it. It's the same thing with AI, the data is in there, you can't remove it. So it's, unless they retrain the model, and then look at specific results, and when they get results, and they look similar to the content they want to remove, they untrain certain portions to remove that data.
I mean, it's so complex, and nobody's going to do that, no multi-billion almost trillion dollar company is going to go and delete your file, it's just not gonna happen. So we're talking about protection and AI generated logos, you know, everything that you talked about, have the actual tools changed to make it easier for you to protect it? Oh, yeah, yeah, they're smarter now. There's AI platforms that are coming out that are only trained using licensed and open source content, so you don't really run the risk of potentially getting into a lawsuit for using the results of an AI model, that's one thing they do.
There's other platforms that are specifically created to maintain confidentiality, because they realize this is an issue, because the public, here's the issue, there's like 80 to 90 percent of the industry is not aware of this issue, like not, because nobody's talking about it, like it's, they're just pushing AI hard. And I get that, I mean, I'm a techie, my background is electrical engineering, I'm a patent attorney, right, I'm an AI law attorney, I do patents, trademarks, copyrights, trade secrets, litigation, prosecution, like I love technology, right, but like I'm not gonna take like a, I don't know if you've ever seen these flashlights that are so strong, you can put them over paper and they burn the paper, right, I'm not gonna give that to like a three-year-old kid, right, and say here's the best flashlight ever, like I know what I have in my hand and then I use it properly. If you don't understand what you're using, your first job is to understand what you're using, like you don't take the assets of your business and all of a sudden just throw them into something because it's fun, like that's just irresponsible, it's not smart business.
So what's the platform that you would recommend, like do you have one or another that's one of your favorites to use? Yeah, I mean, I use, so I'm very different from others though, because I understand the technologies really well, so my system actually, the AI model that I use, it's agentic AI, it's connected to enterprise-level generative AI systems and it's highly confidential and even then I don't put client information in there, I never put client information in there. What I'll do instead is I'll use fake names, right, and fake scenarios and then I'll get the answers and then once I've got the answers, I modify the result to be specific to the client, right, that's the way I use it, but I'm not, like I'm not in a situation where I'm trying to determine the financial well-being of a company by looking at their financial records, like that, so I'm not an accountant, like I'm not a marketing specialist or something like that, so for me it's easier to keep the party out, right, but if you're trying to do an analysis based on your client's spending, you've got to use a confidential version and I would be looking at local AI, I would seriously be looking at some of the available local AI systems that are out there and find one that is safe to use, you know, and so and you're going to start seeing more of these pop up, the reason why is because people are just starting to become familiar with this and so and I just finished writing a book on this and so I guarantee you people are going to start asking questions because I'm making sure that the industry understands this stuff, you know, like there's going to be no more, no more mystery to this, like we're going to make sure they all understand these are issues, these are things you need to be thinking about because, I mean, I've seen presentations by people, again, that are making millions of dollars saying, hey, you know what would be a great idea, you see a website you really like, take the URL, type it into this AI platform, scrape all the images, all the data, like all of the text from it, copy that, paste that into your own funnel or your own website and use that and just change the title, that's called copyright infringement, I mean, that's open shut, like you're dead, like this is not, this isn't even difficult to win, like this is easy stuff and if there's a copyright for it, guess what the damages are, a $65 registration with the copyright office at copyright.gov can get you up to $150,000 per infringement, wow, that's a lot of money, like we're not talking about, oh, well, here I have to pay a $20 fine, like we're not talking a parking ticket, guys, we're talking shutting down your company, wow, so you might want to pay attention and figure out, okay, before I use that AI model, what kind of data am I using, it's public, oh, I'll use whatever I want, I don't care, if instead it's client-related, all right, is it generic client-related information that's not hitting the confidential stuff, as long as my client's okay with it, maybe I'll use the generic stuff, right, now I'm starting to hit confidential information, I got to be looking at enterprise level stuff and making sure that it's not a violation of an NDA, if it's HIPAA stuff, you can go to OpenAI and say, hey, I need a HIPAA compliant version of OpenAI enterprise, they actually have that, it exists, people don't know about it, there's a legal document you have to sign for that and then boom, you get that version and now you can do whatever you want, you can enter all the medical stuff and everything because you purchased the right tool, right, you need the right tool for the job, if you have the wrong tool, you can get into a lot of trouble. And what you're talking about, let's say, for example, a small real estate investor is just getting overwhelmed, just trying to be more effective and so they're gonna spin up an agent, what you're talking about is having something, what they call on-prem and putting a harness on it, on top of that, so that it doesn't go and buy just crazy stuff, rearrange all your everything, so that's really what would be most effective, right? And do you have a preference on, is it Hermes, is it something else that you're implementing? I know there's companies out there like Buddy, Buddy will actually help investors create their agent.
Yeah, I would definitely try working with because Buddy obviously works with me, so if he's trying to figure out if this is safe or not, he'll give me a call. And there's not a lot of guys like me in the industry, honestly, because there's a decent amount of confusion in the legal industry as to what people can do, can't do, and so you have to be willing to be invested in this. For me, I've been writing AI patents for 15 years.
AI just came apart. I know, this is old stuff, guys, this is not new, it's just you've got a really nice new graphical user interface, a new nice API, and so it looks really neat, and now they're pushing it to our consumers more, which, I mean, I've got to admit, it's a lot of fun. I used to tell stories to my kids, and if I knew there was the ability to tell a story, my wife always told me, literally every single night for probably almost 10 years, I would tell a story, and I'd make them up, not like reading a story, I would make them up, and then I had this series that I had, and it was about this ranger who was in the woods, and there was this squirrel that was able to talk, and there was this kid that was lost, and the whole thing, and the storyline just kept going on and on, and my wife would listen in every once in a while, and she's like, how do you come up with these stories? And I'm like, I don't know, I just have a really great imagination.
Now, could you imagine if all of those were recorded, and then the whole series, I just uploaded it into an AI platform and created movies? Yeah, YouTube channels. Yeah, it'd be incredible just to have the series of movies just playing from 10 years worth of stories, and that's why I said I love AI, it's great, it's just you have to train it properly, you have to know what tool you can use and can't use based on the data that you're utilizing, and if you can figure that out, then harness the power and have a great time. I've heard people say AI is a really bad thing, but AI is trained by us, the data, and again, I'm talking about generative energetic AI, I'm not talking about other robotic stuff and everything, I'm just saying in general with regard to the stuff that you and I would typically be utilizing for data, image, video, voice, analysis, and creating of things.
If you train it with the right content, then it can be used for great things. If you train it with the wrong content, it could be used for bad things. And the other thing that I think is really important for you to understand is that the results are not always right.
AI does hallucinate, and so it does provide you with results that look like they're accurate, but they're actually not accurate, and if you utilize them, you can get in trouble. I mean, perfect example of this, there's all the lawyers that are getting sanctioned. There's lawyers that have been sanctioned for $2,500, $10,000, another one lost their license, like they were suspended for a period of time because they were using AI rendered results, court cases that actually weren't real court cases.
But since AI provided the results, they thought that they were real court cases, they provided them to the court within their own court memorandums and their filings, and the judge said, what the heck is this? You're making believe there's case law that doesn't really exist. And there's no, oops, sorry, your honor, like that does not work in a court of law. So they got sanctioned.
So you gotta be careful. What do you like using? I'll be honest, guys, I just use ChatGPT. I think ChatGPT is awesome.
I've used, I've consistently used it, so I know how to use it really well. I know there's a lot of different platforms out there. I just kind of mastered ChatGPT, and so you don't need to go to a billion different other platforms if you can master one.
To that end that you're talking about hallucinations, when I'm creating either marketing campaigns and stuff like that, or I'm actually researching something, I kind of use Claude to check ChatGPT and then ChatGPT to check Claude back and forth. Yeah, you can do that. Yeah, sure.
The other part is, just keep in mind, a lot of people don't know this. There's a number of different AI platforms that are actually trained on the same datasets. Right.
So you might be using a certain AI model to check another AI model, which actually was trained on the same dataset, so it's going to have the same data, so it's not going to help. So it's better to, what I do is I prompt the AI model to provide me with the resources that were used to provide the results, right? So if you're going to push out, let's say as an example, you ask it for research on how many cars were sold in a certain area that came from China, right? And you get the number, and you're about to do a presentation at your business, like the corporation. I would prompt the model to say, please provide me with the specific resources that were used to provide this result, the pages, the URL for the actual article, like literally ask it for where the data came from.
And sometimes you'll see it, it'll say, oh, we're having trouble finding that information. And that's when you question it and say, why did you provide me something that you didn't double check? And then it'll come out with the same thing it always does. I'm sorry, you're right.
I should have blah, blah, blah. That's so frustrating. That doesn't help you.
Yeah, it doesn't help you. So you have to make sure that when you're going after important information that you're going to be presenting on, you're going to be responsible for as a business, ask for where the data came from and ask for the specific URL, the book, the newspaper article, whatever, and get a copy of it and double check it because the stuff unfortunately can be wrong. So when you're using not only these platforms, but you actually have an agent, there still has to be some sort of level of supervision still.
Like you can't just- Absolutely. Yeah. So I have my own system.
I use OpenCloud. I've used that for quite some time. That's the OG.
What was that? That's the OG. It is. Yeah.
Yeah. And it's the one that everybody's afraid of. But if you know what you're doing, you don't really have any problems.
But I have it on its own virtual private network. I have it isolated. It has access to nothing except what I give it.
And I feed it my own legal documents while I've removed the party and removed the specific data. I give it my documents. So it's very familiar with what I need.
And then when I need something, I just tell it, give me blah, blah, blah, like 10 seconds. I've got something that would have taken an associate four hours to write. So it's just remarkable.
But again, you need to figure out what platform is best for you. You have to do the investigation. What data am I using? What legal rights do I need to what's coming out? Do I potentially have any issues with infringement on results? It's your job to dig into that.
If you don't, nobody else is going to dig into it. That's why I wrote the book, man. It's so important for people to understand this stuff as a business.
What are the steps you should be taking before and during use of AI? So let's say I do pay somebody to set up an agent for me. What does a small business owner need to be asking that company about how it's going to go along with me? How reliant I am on them so that I can... Let's say I sell the business or I want to use somebody different than them. How can I make sure that I keep what they set up for me? Right.
So what you want to do is make sure that they assign all right title and interest to the software that they're creating for you over to you. Otherwise you're licensing it from them and they can continue maintaining ownership. The other part is you don't want them using AI to write your own code.
If they're using AI to write your code, then nobody owns it and anybody can do whatever they want with it. So what you could have them do instead is you could have them create it at a modular level, at a top level, mentioning what functionality is being performed and then give that to a coder to write the code and then you own the copyright for that. So that's a way that you can work with third parties who are creating models for you.
The other thing that you need to be careful of is, or at least cognizant of, is the fact that a lot of the AI models that are being created for companies inside really are not truly being created for the company. What they're doing is they're creating something that connects on the back end to something like ChatGPT and you've got your own API like on your computer or you've got this really cool graphical user interface and you've got this really cool website or something and they say, oh, this is your AI, but it's not. It's really connecting on the back end into open AI.
So when you're typing in a prompt or when you're using this website that they created, quote unquote, for you, it's really using an AI tool that already exists. So you need to figure out what you're actually buying. You want an isolated system that's on your PC or on your, that's given to you, not connecting to some third party AI site that's actually doing all the computation and work on the back end, unless you're okay with that.
If you're okay with that, you should be paying exponentially less. You should not be paying a lot of money for somebody to create for you something that looks pretty on a website that where all the ingenuity, all the calculations and everything are done at chat GPT like that, or, you know, Claude or whatever, like that doesn't make any sense. Why would you spend $30,000 for a technology that already exists? It doesn't make sense.
So, you know, if I do set up all these safeguards on my laptop, let's say I'm on my, on my hard, on my computer, I set up all the guardrails, all that kind of stuff. When I'm using chat GPT, when I log into my phone at chat GPT, do all those come with it? Or do I need to make sure that I set it up on my, on my phone also? Yeah, it's a good question. I would double check it.
It should, the settings that you do on your PC should move over to your, like your cell phone version, because it's connecting to a remote server and the server should have all your configurations. But I would just double check. I just, personally, it doesn't hurt anything and it only takes a second.
I mean, if I even walk people through right now, if I open up, let me see, where's my chat GPT? There it is. Chat GPT. I click on the button.
All right. So when you open chat GPT, you'll get, let's see if people can see it all the way up, up here, there's like these dots, these lines, right? So on the left side, you click on the lines or when you click on the lines, there's a menu, right? It has all sorts of options. If you scroll down, see all.
So you can check all the settings that you set up. Oh yeah. Yeah.
What I do is I, I click on, and I'm going to show you guys how I do it on mine, which is easier than doing that. So when I click on the two dots down here, there's the, the icon. Do that again.
There we go. There's a little circle there that says PN, right? And you'll have your own with your own initials. It's right there, right? So click on that.
When you click on that under the options, when you scroll up, you'll see one option that says privacy center. Uh, it's right there. So you click on privacy center and you click on privacy center.
You'll see the option that says data, what is it? Data use and access. That's it. Data use and access.
You click on that and now you'll see a bunch of options. One of them says model improvement and you click on model improvement is a big fat button that says manage. Yeah.
You click on manage and the very top of it says improve the model for everyone. That thing should be off. There's no reason for you to have that thing on.
If you have it on you, all the stuff that you've been entering is public. So that's how you turn off, you know, the, and that, that took like, what? Yeah. 10 seconds.
So just check. And this is the same thing for Claude. It's the same thing for all these other models.
They all have ways to turn off training of the model. You don't need, you don't need to be giving data to the company that you're paying. Like they are, they're benefiting from the millions of people that are using their sites that are allowing the data to train their model.
It's making their model better. It's making it more up to date. It's giving them more information for calculating.
Uh, and they're not paying you for it. Especially if you consider yourself one of these, you know, really curious or creative people, you know, that's, that's the last thing that we got to hang on to, right? It's like curiosity and, and what keeps us human. Uh, Peter, we're, we're, we're getting close to wrapping it up.
I, I want to ask you one question. Uh, what is one prediction that you can give us that at tsunami three, we're going to be like, all right, Peter, did this come true? Uh, I think the copyright, well, once tsunami three, what's the date March, maybe not by then. Okay.
So let's say a prediction for 2027. How about that? All right. All right.
For 2027, I think the copyright office is going to come out and make a statement saying that certain things created by AI are protectable under copyright law. Interesting. I think that's going to happen.
And that's going to change the industry. That's going to be a big deal because people are writing prompts that, that make it so that the AI platform, the software, the model, basically it's not leaving a lot of creating to AI. It's actually basically what it's doing is rendering exactly what you said, right? So if you're writing a prompt to the point where you really are the creator and AI is just, is just rendering exactly what you're saying, I would argue that you should get a copyright for that.
Right. So as an example, if you say, give me a frog, the color of the skin is this hex number. The eyes are exactly this dimension.
It's located at this spot. It's this number of toes, the color of the toes are this, like, I mean, you go into such great detail that really there's no creating by AI. It's just rendering.
I think the copyright office is going to come out and provide situations where they're going to give you copyrights. Now that's going to change the industry. Meaning if you get prompt engineers who know how to do that well, now, all of a sudden, the things that coming out of AI are protectable.
Now you've got people that can at high volume, take over an industry and own. I mean, imagine like, I don't think this is going to happen. I'm just saying, this is an extreme example.
So you can see what kind of things I'm talking about. Imagine all the sunsets, all the sunset paintings that you could ever think of, somebody owns them. Like, just think of that.
Right now that that's not going to happen. I'm just saying, imagine a case where somebody literally is running and filing copyrights for every, every thought you could ever have associated with sunset. And now anytime you paint something, you get sued.
Right now, that's not going to happen. I'm just being clear. That's not, not at that level.
It's not going to happen. That's an extreme situation, but it does show you that companies will be able to take advantage of technology if they can figure out the way to own what's created using AI. And so it's, it's, I think in 2027, you're going to see some moves in that area.
All right. You've mentioned it a couple of times in the interview and, you know, we, I saw it at Tsunami, too. You wrote a book.
What made you, what's it called? And what made you want to write it now? Yeah. So this is the book here. It's AI Side Effects May Include.
This book, this book not only wakes you up to the issues, but it shows you how to get around them. It shows you exactly as a business, what you should be doing to be highly successful in utilizing AI, not losing your content, not getting sued for using results and actually being able to own what you enhance using AI. Plus it gives you at the end, I actually lay out all the parts of an AI business plan.
Every business should have an AI business plan, which lays out what AI is, what the risks are, which models you're allowed to use, which ones you shouldn't be using. Who is the point person at the company who's responsible for AI related questions, laying out for vendors, are they allowed to use AI in rendering work or not? If they should, if they shouldn't, like all that stuff is in the AI business plan. And at the end of this book, I actually give you all the chapters that are in my own AI business plan that I charge for.
And so that you guys don't know what actually belongs in there, but it hits everything from vibe coding, to writing books, to customer, I mean, everything's in it. And so it's basically the last four years of training that I've done to over 50,000 people on what I call the AI do's and don'ts masterclass. And so this book makes it very easy to figure out what you can do and can't do and how to do it.
And so it's at, if you go to AISideEffectsMayInclude.com, it's there. But it'll be on Amazon, like very shortly. And on Amazon, it'll be $24.95. But on this other site for now, I think it's the $24.95 plus shipping.
But on this other side, I'm just putting it at like $19.95 and I'll just mail it to people. It sounds like it's the AI Bible for entrepreneurs, basically. That's what I'm hearing.
Literally, some of the people that have read this are just going, what the heck? You're showing people actually, if they're using AI to create a logo, you're showing them how to check if it potentially infringes somebody else's.

