Episode 20 - AI Compliance & IP Strategies: Essential Insights for Modern Businesses

In Season 2, episode 5 of Artificial Antics, host Mike Onslow sits down with Vincent Allen, an intellectual property attorney, to explorer the essential strategies for AI compliance and innovation protection in today's business landscape. Discover how to safeguard your creations and navigate the legal intricacies of AI technology.

Vincent shares his journey from engineering to patent law, highlighting how businesses can leverage AI safely while maintaining their intellectual property rights. From discussing prompt best practices to sharing real-life hallucination horror stories, this episode is packed with insights into avoiding legal pitfalls in the AI era.

Vincent also provides tips for businesses utilizing AI technologies and the importance of having robust AI policies in place.

Contact Vincent Allen:
- Email: allen@caglaw.com
- Phone: 972-367-2001
- Websites: https://caglaw.com, https://aitriagecenter.com

Tune in to this episode for invaluable advice on staying compliant and innovative with AI.

#AICompliance #IPLaw #BusinessInnovation #ArtificialAntics

Stay Updated:
- Subscribe to our channel for more episodes like this: https://www.youtube.com/channel/UCXz1ADq4dDQ5yPy40GWDepQ?view_as=subscriber&sub_confirmation=1
- Follow Us: https://www.buzzsprout.com/2178171/follow

Connect with Us:
🌍 Artificial Antics Podcast Website: https://antics.tv/
📰 AI Bytes Newsletter: https://artificialantics.beehiiv.com/
🚀 Artificial Antics Business Site: https://artificialantics.ai/
💼 LinkedIn:   https://www.linkedin.com/in/artificialantics/
🐦 Twitter: https://x.com/anticslab
🍏 Apple Podcasts: https://podcasts.apple.com/us/podcast/artificial-antics/id1694398244
🎧 Spotify: https://open.spotify.com/show/0QpTQhJrdXNeqc5TX8clmj

Transcript

**Natasha:** [00:00:00] Welcome back to Artificial Antics, where Ricoh and Mike will talk about the implications and opportunities around artificial intelligence, machine learning, and deep learning. On this episode, Mike talks to Vince and Allen about protecting your creations in the AI era, covering prompt best practices, real hallucination, horror stories, and must know patent strategies.

Tune in to stay creative without courting legal trouble.

**Mike Onslow:** everyone, this is Mike from Artificial Antics. We've got our first show in a while here. I'm joined by Vincent Allen, who is an intellectual property attorney and, who's been getting really into the AI technology space. As you know, you can imagine there's lots of, interesting stuff to talk about there.

So, without further ado, I'd love to introduce Vincent yeah, Vincent. So, how you doing today?

**Vincent Allen:** I am doing great. How are you, Mike?

**Mike Onslow:** I'm, I'm doing great. Like I was saying, it's a, it's a sunny day outside, it's [00:01:00] Friday, and looking forward to, to relaxing with the family this weekend.

**Vincent Allen:** You and me both.

**Mike Onslow:** So Vincent, you know, maybe as a starting point, why don't you go ahead and introduce yourself, introduce your firm, give the listeners a little bit of color and background on you.

And especially I'd love to hear, about your AI journey, right? So, kind of how all this came to be and why we're talking today.

**Vincent Allen:** Sure, no problem. , well, I'm a managing partner now at a intellectual property and technology law firm called Carstens Allen and Gourley. And, and we're up in the, Dallas area, north of Dallas in Plano is office now. We've been in this area, over 25 years, and in fact, I have been with the firm about that long as well.

, I started out right outta high school. I loved aviation. , you can see I've got the airplane [00:02:00] in the background there. And so I went to aircraft maintenance school and. Thought I was gonna be a an AMP technician, but ultimately fell in love with the electrical classes that I was taking and decided to go on to electrical engineering.

And so I did that and in the process of, getting ready to graduate, one of my professors said, you know what, instead of going to work, you ought to, come get your master's in electrical engineering and. You can work on this project that I have going be my graduate assistant. And I said, well let me, let me think about that.

I never really thought about, you know, continuing my education beyond undergrad. I was just gonna go to work. So I stumbled across patent law when I was doing some research online and there was a big demand, for electrical [00:03:00] engineers to go into patent law. And the more I read about it, the more I felt like that would be a good opportunity for me.

And so ultimately, I, I applied to law school. I went to Baylor Law in Waco, and then ended, ended up here in Dallas and have been here ever since. during the course of my career, I've done a lot of, patent and trademark filings, including both the procurement of the intellectual property as well as litigating patent and trademark disputes that might arise.

And during that time, I've even, you know, occasionally would address artificial intelligence in the patent applications. It's not a new technology. It's been around for a long time. But it didn't become super popular until the large language model came out at ChatGPT in particular, back in November of [00:04:00] 2022.

And at that time I started playing around with ChatGPT and I realized it was gonna be pretty revolutionary. And I said I needed to. Understand, you know, how it works and, and come up to speed on some of the legal issues that might be present. And one of the things that I identified pretty early on was the problem with putting in confidential information into your prompts or using it, in order to generate a result from ChatGPT.

And the problem is that it, the model can be trained on the prompts that are put in. And so you risk. exposing your confidential information if you put it in, with the original ChatGPT. , now there's some, if you get an enterprise version or if you have, click a button, you can try to avoid the use of your data for training.

, [00:05:00] but that was a risk that I identified and, and so I said I'm gonna start doing some talks about, you know, the potential risk associated with using. Gen AI and I started that in early 2023. , gave a couple of presentations and then I was getting ready for one presentation, with the Plano Chamber of Commerce on the state of ai.

, included some speakers from, you know, Microsoft and Google as well, talking about their technology. And of course, I'm. I'm kind of the naysayer, not not a naysayer, but I'm kind of the guy that has to rain on people's parades and say, look, you know, you gotta be careful about what you're 

**Mike Onslow:** Right. 

**Vincent Allen:** doing. And so anyway, I did that and, and also I, I met with a gentleman named Rob Taylor before the event, and I found out he had been vetting.

Many use [00:06:00] cases already for Gen AI. And , so we decided to partner up and he joined our firm. We started up what we're calling the AI Triage Center, and that's what we used to describe, you know, all the services that we provide related to Gen AI, from the initial development of the software or the products that you might be implementing to.

Developing agreements, consent forms, etc., that you might need with the application of Gen AI. And then also, protecting the IP associated with it. So if you need to file a patent application on the technology that you have developed, even if it might be based on a, you know, a working model, I mean a somebody else's model like open AI's model.

You can still get patent protection on the way that you're using it. And so we advise clients, in that [00:07:00] area and help them secure, intellectual property as well as stay outta hot water. You know, when it comes to the compliance issues.

**Mike Onslow:** Yeah. No, that, that's great. All all really important stuff. You know, one of the things that I was thinking about with, you mentioned agreements, right? And different things. And I said, gosh, so here, here's one that I don't even think I have on the list of things to ask you about, but I think it might be interesting is, you know, a lot of people these days I think are using tools like Chat, GBT and other LLMs to generate their.

You know, operating agreements and a bunch of other stuff. Have you any stories on either side where, where you've seen, a, a big benefit or a big, you know, downside or let's say, you know, big issues with that? Like any, any kind of thoughts on that?

**Vincent Allen:** Well, you say operating agreement. I mean, that is a, that's a legal [00:08:00] document in my world. You know, an operating agreement to, to govern how A LLC might be run, so I think you need to be really careful. In fact, I'm telling you not to use, you know, ChatGPT or other LLM to create legal documents. Now that doesn't mean you can't ask it questions and try to, you know, get the answer to certain things, but I think it's really important to, to also consult a lawyer in that area and make sure you're doing it right.

You know, even if you come up with a first draft of, of the agreement. , those chances are you've missed something, or chat. GPT missed something. And so if you're not careful, you can really get yourself in, in hot water using it for legal agreements. But at the same time, I do think it's a very powerful tool.

I think if you use the right prompts, [00:09:00] you can get some pretty good results out of it.

**Mike Onslow:** Yeah, that, that makes sense. , and, and I would say. You know, like Vincent says here, I, I think that at the very least, right, it's, it's, Hey, maybe it generates a first draft, gives you some ideation. , you should always, always, always have that reviewed by, you know, a, a a law firm professional lawyer.

I mean, it's, it, it's just too important, right? And, as we know, as probably all the listeners know, and definitely, you know, Vincent. It hallucinates, right? It either, it might miss something. It it, it hallucinates it, you know, it does all kinds of things. And I remember one of the early episodes that Rico and I did, there was a lawyer who got in a lot of trouble because he cited a case that didn't actually happen.

And of course he said, is this a real case to Chakin? And said, yeah, this is a real case. And so, so he blindly

**Vincent Allen:** I. [00:10:00] Exactly. So the, the judge in that case, and they were, there were actually several case citations. And the problem that the attorney ran into is, he used it. He didn't check the work to make sure that the cases were actually good cases, and that it turned out that ChatGPT just made up the cases.

They weren't even real cases, but they looked real because they were cited the way that legal cases are cited in briefs. But it came up in, at, in, in front of the judge and the judge. You know, asked him, you know, did he check these cases? And he's like, yes, I checked those cases. Well, he got himself sanctioned because they found out that the cases were not real and that he had used ChatGPT and then lied to the judge about it.

**Mike Onslow:** Yeah, that's, yeah, so I mean, that, that's the thing, [00:11:00] right? Like from from the start, I, we, we've been informing folks like, hey, don't, don't ever blindly, even if it's not a legal document, right? Even if it's just, you know, research for school or whatever it is there, you know, you need to be verifying.

These things, right? , and, there's a, there's a nomenclature called 10 80 10, which is 10. You're setting the vision, you're getting an initial prompt set up, then 80%, right? Like the, the, let, let ChatGPT or whatever tool you're using, do some work for you, but then that last 10%, right? That's where you're scrutinizing.

You need to really be sure that it's the right stuff and sometimes. It's back to square one 'cause it's very, as you know, Vincent, it's very iterative, right? Like getting the right answer out of these tools. The first time I. Can happen, but it's a, I I'd say it's a very, it's a very low percent. So folks out there, you know, make sure you're iterating, make sure you're checking and [00:12:00] scrutinizing and, yeah.

So that, so that, yeah. No, that, that case was really interesting and I'm sure since then, right. I'm sure you probably don't hear about him as much. 'cause that was like the first time I think it was really like everybody was like, oh, you know, there's probably been a lot of that, because I'm sure he's not the only one who's ever done that.

Right. So.

**Vincent Allen:** Well, I, I haven't heard of another case where they, you know, mis sighted, cases. But, but I think because the reason being is he, you know, he probably used it and didn't realize that ChatGPT Hallucinates. But once that case came out and every, all the attorneys found out about it, you know, they.

, being cautious about it, you know, how they're using it. And also even some of the judges are placing restrictions on, briefing and, and having attorneys, sign a certification that they either didn't use [00:13:00] ChatGPT, or they did use it and they've checked everything. 

**Mike Onslow:** ha. 

**Vincent Allen:** So yeah, and that was actually Judge Starr here in the Northern District of Texas did that.

And, and there's been some others who've, you know, implemented some local rules for that purpose as well.

**Mike Onslow:** Yeah. No, that's, that's, that's good. I mean, like, it's, it's like fool me once type stuff. Right. So, so as far as, you know, let's, let's kind of dig into intellectual property law, right? And AI and, Can you give us just some you know, idea of, IP law in general and like some of the nuances there, and then really how ai, the, the, as you said, it's not a new thing, right?

AI's not new, but I think, let's call it modern AI, like LLMs is definitely newer. And, just kind of give a, give the listeners an idea of. some of the stuff that, would fall into that realm of intellectual property and then fall into the, realm of intellectual property, [00:14:00] specifically around, AI creations and inventions.

**Vincent Allen:** Sure. So, you know, there's several aspects of, or issues I should say that come up with respect to intellectual property and artificial intelligence. , one of the issues is, can you protect the inventions or the works that are created using a tool like. ChatGPT or Mid Journey, whatever it might be. Are you able to protect that?

And the copyright office and the patent office, both, in my opinion, are probably a little more cynical, I guess, of allowing protection for. Works that are generated by Gen AI or inventions that are created using Gen AI. So let's talk about patents in particular. In order to be an inventor, you have to conceive an [00:15:00] idea, and you can't be an inventor if you're a machine or if you're an animal and you're not a person.

The law is pretty clear that you need to be a person in order to be an inventor. However, what the, what the patent office has said is you can use Gen AI tools or other AI tools to assist you in creating inventions. As long as you have a substantial contribution to the actual creative or conception of that invention.

It if you didn't, if you didn't conceive the idea and you just watched Gen AI produce it for you without any input, then you're not an inventor. So compare that though to the case where, um. An inventor, the inventor of the microwave oven, for example. [00:16:00] He discovered it by a pure accident. He had a candy bar in his pocket while he was testing out, microwave transmissions from a magnetron. And he discovered that his candy bar was melting.

**Mike Onslow:** Wow.

**Vincent Allen:** when he discovered that he put some other food, he saw that popcorn would pop when he. You know, put it in these microwaves and, and so that's how the microwave oven was invented. It was an accident. Now he accidentally discovered it, but he did appreciate the fact that he had just discovered a really clever invention.

And so he ultimately obtained a patent on the microwave oven. But again, he's an accidental inventor. So how is he different? My question is how is he different than the inventor who uses AI and accidentally discovers an output from [00:17:00] AI that also is patentable? , I think that the patent office is probably a little too restrictive in requiring that substantial contribution, to the conception of the idea.

And so then the, the next issue is with regard to copyright law. , under copyright law you have to be an author, of. The work that is being generated and the copyright office says, if, if it's only Gen AI, that is producing an output, and you make no changes to that output, then you're not an author.

You can't get copyright protection. It doesn't matter what kind of prompts you use to generate that

**Mike Onslow:** Okay.

**Vincent Allen:** according to the copyright office. Their opinion is that, well, you don't really [00:18:00] choose how Gen AI expresses that idea. Gen AI does that by itself. And, and so I, I think that's problematic too because what's the difference between Gen AI and, you know, putting a prompt in and getting an output out, out?

What's the difference between that and somebody taking a picture using a camera? You know, they didn't actually make that picture or form that film, or negative the camera did and they just simply pushed the button. Now obviously photographs a copyrightable. There was a Supreme Court case that it said that you can get copyright protection, but just like Gen AI now, back then there were people who didn't want protection for photograph because they didn't view it as a really [00:19:00] creative avenue of expression. And so therefore, you know, the painters and. , they didn't want

**Mike Onslow:** Yeah,

**Vincent Allen:** be protected because they felt like it was gonna take away the market for their, their works.

You know, their paintings that they did, portraits, you know, of people would, would no longer be needed because people can just, you know, use this cheap, easy photographer to take a picture of 'em.

**Mike Onslow:** Wow. When now, now when was that? Because I, you know, I, I don't know, I don't have a point of reference for around, but like, do you know, like how long ago was that, that, that, that type of like the camera was in, invented? Was that I, I, I don't have a good point of reference. You probably know, but I'm, I'm super interested in that now.

**Vincent Allen:** I think it was in, the 18 hundreds. I, I, I can't remember the exact year of that, but, um. , I would have, I would have [00:20:00] to get that information and get back to you. I don't have that in front of me right now.

**Mike Onslow:** No, that's fine. It's the same idea, for, for a bunch of different things. Right? Like one of the, one of the stories that I, that I heard, that was kind of interesting, was, you know, like if you look at the different things like, you know, Photoshop, right? I remember Photoshop. You know, it was like, oh man, you know, graphic design.

And people were like, oh, it's print is whatever, and this makes it too easy. And then think about this, right? Photoshop's plugins as well. They came out with those like Kais power tools and it was like, oh, you're not a real Photoshopper, you're just using the plugins. So I, I do feel like this stuff ends up getting layered, especially in the artistic community.

This stuff just keeps getting, it continues to get layered and it always will, right? Like, think about. Think about how we used to produce music, 200 years ago and how we produce it now. It's a whole, it's a whole different thing, right? And, and [00:21:00] you know, it, it's based really on the capabilities, not necessarily of a person, but of. technology in general, what capabilities are out there? Oh, well if you can do this, then why would you do that type thing? It's sort of like doing, you know, math on a piece of paper, which I think is important to some degree. But then there was a point where the calculator came out and there's not, there's not a reason to do that.

Like, alright, now we've gotta do every piece of math right on a, on a piece of paper, long hand. So we, that's brought us up to like, okay, let's. Let's think about higher level things and more relational things and whatever it is, right? So that's kind of my take on it.

**Vincent Allen:** Yeah, I, I agree with that. I, and I think that, and, and one example of a case that I, well, I think the copyright office got it wrong, is the Jason Allen case. And that case is, you know, it's not just a case of putting one prompt in or one question, you know, give me a image of [00:22:00] an opera singer, right? That's not what he did.

he took, I think it was 764 prompts that he used. He had this vision in his head about an image that he wanted to create. And he kept iterating. And that's what artists do. They iterate until get something that they're comfortable with, that they like, that represents the vision that they were trying to express.

And so that's what he did. 624 times, he'd get four images from each one. He picked the one that he liked the best, and then continue with that one. And, and then at the end of it, he used, um. Adobe as well as a, resolution improvement tool. And he made some changes to it to, to get the end product.

And then he submitted that to the Colorado State Fair and actually won in his category. Yeah. But then they found out that he used [00:23:00] Midjourney to create the image, and they didn't like that. So they, they created a new category and had some rules that required disclosure of the use of Gen AI before he submitted, but in the meantime, he filed a copyright application, but the copyright office refused to grant him.

A copyright on the entire work. He, they said he had to disclaim the work that was generated by Gen AI. They would give him a copyright in the changes that he made, but he wasn't happy with that. He wanted to be able to protect it all, and so, so he filed a, a suit in Denver. District court to appeal, that decision by the copyright office and that that's pending there now.

**Mike Onslow:** Okay. Yeah. I, you know, and, and one of the things now, now that's, there's [00:24:00] now kind of a trend right, in what we're talking about, which is businesses and really, pretty much anybody right now has to adjust how they operate. Based on AI technology and the differences, right? We talked about law, right?

There have to be specific rules put in place and a framework put in place to disclose, right? That's a big one. And then similar to this artwork thing, right? Like it's, you know, you talked about, hey, he wa he won the competition, then they found it out that they said. Well, let's have another category. Now, I don't necessarily agree or disagree with that.

I probably come on the side of Disagreeance, to be honest, but I, you know, it's something that was adjusted based on, you know, based on, a new finding, right? Which is that, AI art can win contests. I think there was another one, right, like a Sony. Art Fest or something where they, you know, the [00:25:00] guy who submitted it, it was a, it was a digital piece of art that was also generated by, ChatGPT or one, or not, probably not ChatGPT, but something like Midjourney.

Right. And he won that. And then he, you know, of course he, he, he said, oh yeah, that, that was AI or whatever. And then, you know, people freaked out. So,

**Vincent Allen:** Mm-hmm.

**Mike Onslow:** So, yeah, I think the world is definitely adjusting and changing all the time and we're, we're only gonna see more of that right. As we, as we come into, you know, future years here.

**Vincent Allen:** Yeah, I, I agree. And I, I think, you know, a lot of it is, you know, people are afraid or have fear, you know, of the unknown, and so they're not really sure. You know, how AI should be treated, similar to the camera by the way, that, that case was an 1884 decision by the, Supreme Court of the United States.

, it was a, it was a photograph of Oscar Wild. And, [00:26:00] anyway, they, prevailed, but, but at similar situation at that time, there was a lot of, um. Uncertainty about whether this should be protected or not. But you know, ultimately the Supreme Court found that there was a, you know, creative expression.

And I think similarly here, the courts are likely to find that even the use of props. , to create something using Gen AI is enough to rise to that level of creativity that, you know. The interesting thing is, is that the, the Copyright office says, well, you can get copyright protection in your prompts. Well, that's not necessarily what I want. I mean, maybe I do want it, but I wanna be able to protect my resulting work, because if I can't protect it, then that means anybody can copy it, right? And so I think you have to kind of balance that. Are you going to reward, [00:27:00] you know, creation of different works using Gen AI?

Does it really matter that it's a new tool that's been developed that's, you know, pretty well advanced, but it, it's not all that different than using Adobe Photoshop. You know that in fact, Adobe now has some AI tools,

**Mike Onslow:** they do.

**Vincent Allen:** in it as well,

**Mike Onslow:** Yeah. Yeah. No, a absolutely. So, you know, one of the things that I also wanted to kind of chat with you about, because I, I, our listeners, a lot of them are in the business world. We have a lot of, you know, CEOs, founders that, you know, they're running a business and they. They know that, there may be rogue AI being used, you know, not, not maliciously, right, but they just have no idea, you know, what's going into these and different stuff like that.

So, can we talk a little bit about, you know, you, you kind of alluded to it in the beginning of the show, right? That compliance and some of those considerations for, [00:28:00] you know, common, you know, like, like PII is a big thing and stuff like that. Can you kind of talk to like. The, some of the tips you would give businesses in general as, you know, as the advent of, LLMs and other tools, become just available to anybody and people know about them.

**Vincent Allen:** Sure. Yeah. So you know, there's two different Ways you can look at AI. One being from the development perspective. Am I, am I a startup or a business who's developing some new technology that uses AI? And if so, how do I make sure that I produce the best product that I can that's not gonna get me sued later in litigation by my customers or their customers?

So that's, that's one aspect that we can look at is, you know, actually how do you develop it in a responsible manner, so that you get good results out of it so [00:29:00] that it doesn't have bias. , and so that you know, you're not gonna have these claims, you know, how do you oversee, that product and make sure that it's acting like it's supposed to be acting.

And, and not going rogue. Right? And then the other approach or consideration is, well, what if I'm just a business who wants to use AI technology? I'm not developing it, but I wanna use somebody else's software, it's been said, well, maybe you just need to get an indemnity provision and get.

The software developer that provides you with the software to indemnify you for any bad results. Well, it, that's, that's a good idea. But it doesn't absolve you from liability if you choose a product that's not really good for what you're asking it to do. So what are some of the things that can come into [00:30:00] play?

And, you know, it, it all depends on the use case. You know, how you using it some. Issues of Gen AI or, you know, much higher risk compared to others. And so you have to take that into consideration. But obviously one of the things you mentioned was the PII and you know, the security of the data that you're putting into the model, you have to make sure that you've got it structured, in a way that's gonna protect that, that.

You have, you have permission to use whatever data you're using, and that you, you've got the agreements in place, for that, that you're securing it, you know, behind the firewall or, you know, it's not getting out to train. One of the large language models showed that it, it would inadvertently be disclosed to someone else.

So you have to worry about those sort of considerations as well. [00:31:00] But then it, again, I'm the one using the model, so I need to make sure it worked for what I want it to work for. I need to make sure that it has the proper disclaimers in place. , depending on whatever regulations might be a applicable to that situation.

So does it have this, the, consents required for your customers to sign when they use it? , that they acknowledge that this information is being used to process or, or being used by, at a third party, to process their data and give whatever. Output you're asking it to give. So the consumer needs to be made aware because if you don't make them aware, then you may be subject to lawsuits.

And one such lawsuit is the Patagonia case. And in, in that [00:32:00] case, Patagonia was using a third party to listen in on, customer phone calls. And process the data from the call using AI and then give some recommendation to the customer service representative based on that. Well, there's been a class action lawsuit filed claiming that Patagonia didn't give the appropriate notices that their data was being used by AI.

They, they simply stated that the typical statement that your call may be recorded. For quality assurance purposes. , but the cla the class action saying, well, that wasn't sufficient. So those are some of the things that, you know, you might, need to take into consideration. But again, it depends on your particular use case.

And if it's a higher risk use case, like HR and employment where you're making employment [00:33:00] decisions or you're giving employees information about their benefits. That kind of thing, then you, you definitely need to, you know, put a closer eye on how that works and, and also provide the oversight necessary.

Do you need a human in the loop, for example, to check the results that you get out of it

**Mike Onslow:** Yeah. A absolutely. Now, with that Patagonia case, is that still ongoing or did they actually, was, was there a ruling in that one?

**Vincent Allen:** there? Um. I, I don't think there's been any, final determination in the Patagonia case. , so it's still pending as far as I'm aware.

**Mike Onslow:** Yeah. Yeah, that, that's really an interesting one because I know a lot of businesses I. , I'd say pretty much every business right, has that Your call may be recorded. But, but yeah, having those additional terms, right? And, and, you know, consents from people. [00:34:00] One of the, so I, I'm in the communications industry and that's a huge thing, right?

The technology that we're spinning up, right? It all has that consent, the trust built in. there's a real, implication right. Of, of getting that wrong. So, so tell me a little bit about what, what you and your business do, because I believe you actually could help businesses with that, 

**Vincent Allen:** Sure. Yeah. As I mentioned, my, my colleague. Rob Taylor, has vetted, many different use cases for, for a, you know, large, fortune 50 company, but over the course of, I think it's over 200 cases now, but essentially what we do is we look at how the model or how the system works, the software look under the hood.

You know, look at the prompts that are being used so you, so that we can ask questions about, well, why are you using this [00:35:00] particular prompt? You know, isn't that likely to give you an erroneous output or something that might be biased? For example, that, and so we do that kind of thing. We also look at, You know, the model that's being used and the terms and conditions associated with that particular model. So for example, we would tell you to never use deep seek,

Are 

**Mike Onslow:** right.

**Vincent Allen:** the, risk that your data gets, shared with the Chinese government. But you do need the, the terms and conditions of different models are different.

So some of them require you to indemnify the model owner for any liability that results. And others say, okay, well we will indemnify you, for example, for any intellectual property infringement claims that arise, because there's just whole another set of lawsuits [00:36:00] involving fair use and whether it's property used your, copyrighted materials without permission to train a model.

So, OpenAI is, pursuing that and, you know, we we're gonna find out whether that's permissible or not. But, so that's just one example of, of something you may wanna, wanna look at. And then of course, you know, the contractual terms with your vendors, making sure that those are appropriate. , whenever you are developing or, or.

They're helping you to develop an AI solution, knowing who's gonna be responsible for what, because again, everybody can ultimately be held liable. In a lawsuit, the developer can be held liable. The user of the software can be liable because they're the ones that chose it. It, it's, it's no different than saying, okay, I'm gonna hire this employee for my [00:37:00] company to handle HR.

And then that person, you know, makes a, a biased hiring And, and now you're accused of discrimination. because that employee worked on your behalf. , well it's no different to use software to work on your behalf than it is to have an employee to work on your behalf. So you have, you have potential liability there as the company that you need to make sure you protect against edge.

**Mike Onslow:** Yeah. No, no, absolutely. Yeah, so how, so, so as we start to wind down here, Vincent, a couple of things. One is, any, any other tips for, business owners, to make sure that their business is protected, as they, as they start to use AI and their people start to use AI?

**Vincent Allen:** Well, so one of the common ways that, AI is used is to generate content, whether it be text or artwork, and what I would say, at least under the current state of the law. [00:38:00] I would recommend that you make changes to that show that. You will have copyright protection in the resulting work that as long as you, you're adding some level of creativity to the document by, you know, revising the wording of the text or.

Revising the look of the image that has been created. So if you can do that, then that'll help you get copyright protection. You know, if you're actually developing software, and even if you're using a large language model off the shelf, keep in mind that you may be able to get, patent protection on that particular application that you've developed.

If you think it's new and non-obvious, in view of the priority. And then of course, making sure you have the appropriate policies in place, within the company to protect yourself. Make sure everybody knows that they can't just put all this information into, you know, the free version of [00:39:00] ChatGPT,

**Mike Onslow:** right. Yeah,

Yeah 

**Vincent Allen:** And, and that kind of thing.

Don't assume that everybody knows. As much as we do that, you know, we are in it every day. Not everybody has that same experience.

**Mike Onslow:** Yeah, that's right. That, that's the first thing. When, when I, when I really took up the hey on AI had at uh. At Clarity and like I've gotta figure this stuff, stuff out and make sure we're using and adopting it both efficiently, but like responsibly too. Literally. The first thing that that I did is I, formed A committee, I called 'em the AI Ambassador group, and we built out. The policies, the, one of our people call, one of our directors called it the do not do list. Right? Like those different things. And then it's not, it's not good enough to just have it somewhere in your knowledge base or your handbook. We, we do trainings, like very frequent trainings like out in the open with the whole company.

You know, here's what PII looks like. It's not just [00:40:00] social security numbers, it's other sets of data that can be used to identify somebody. Right? So, so yeah, it's super important that that one I think is like a one 100 level. Like every company needs to have at least an AI policy, in place. And, and so, yeah, no, that's a great one.

And, and yeah, so, so Vincent, how can people get ahold of you? Because I would say that, there will be listeners that are interested in, you know, especially I think that side of building a software product and really understanding and having, from a, from a legal level, right? An advisor to go through and the prompt reviewing the prompts.

That's one of the coolest things I've heard in a long time, honestly. So how can folks get ahold of you and, and your firm to help with that?

**Vincent Allen:** , Uh, sure. My phone number is 972-367-2001, and my email is [00:41:00] allen@caglaw.com, and you can also find our website at caglaw.com and you can read more about the AI Triage Center at: aitriagecenter.com.

**Mike Onslow:** Oh, awesome. So you guys actually have a, a site spun up for that. I'm gonna go check that out myself after, after the show here. And folks, we will have those, that information to get ahold of Vincent here, in the show notes and, yeah. Vincent, I just wanna thank you for being on the show.

, really glad we got together. 

**Vincent Allen:** Yeah, so I, I certainly appreciate the opportunity to, to talk, Mike. I really enjoyed it. You have a, a great way about, you know, asking questions and, I think we did a lot of good here today.

**Mike Onslow:** Yeah, absolutely Vincent. So folks, check out Vincent's site. , especially if you're building software products, as I know a lot of the people in the community are, super important to, to be broaching these and not just, not just, skimming [00:42:00] over the, ah, well, you know, the bias, it, it's not a big deal or whatever.

Prompts, prompts have a lot of power in 'em, and they need to be structured correctly, especially if you're gonna then. , bring that out to, to other people. And, yeah. So that's it for this episode folks, and, we'll see you back in the lab soon. Thanks, Vincent.

**Vincent Allen:** See you Mike.

​