AI and Intellectual Property: The Legal Risks, with Erin Austin

How this transcript was made

Transcribed by machine from the published audio, then corrected by hand. It is not a verbatim record: the changes below were made, and are counted.

This episode

87 speaker turns, 78 changes.

• closing CTA identified as CTA 1 - Share (match 100%), replaced with the authored script and attributed to Max

What was changed in these transcripts, and what was not

Generated: 2026-09-20 from 2026-09-20-named/, itself transcribed from the audio that is live on the feed. Every change below is mechanical and counted. No sentence was reworded, no claim altered, nothing removed for being awkward.

The editorial decision

Owner's choice, 2026-09-20: light clean. "Um", "uh" and stammered word repeats ("and, and") are removed, because nobody intends to say them. "You know", "I mean", "kind of" and "sort of" are KEPT - they run to roughly 2,300 instances across the catalogue and stripping them would make the page read as somebody else.

Counts across the whole set

Change	Count
stammered repeat removed	667
um / uh removed	654
space before punctuation	241
split contraction	132
guest turn named (Joe Casabona)	101
organization -> organisation	71
diarisation fragment returned to the host	62
ai -> AI	57
empty turn dropped (was filler only)	36
synthetic outro restored from script	21
organizations -> organisations	18
hyphen followed by space	14
realize -> realise	14
A.I. -> AI	10
recognized -> recognised	9
gray -> grey	9
summarize -> summarise	9
color -> colour	9
behavior -> behaviour	8
synthetic intro restored from script	8
analyze -> analyse	7
behavioral -> behavioural	7
recognize -> recognise	6
percent -> %	5
personalized -> personalised	5
skeptical -> sceptical	5
analyzing -> analysing	5
center -> centre	5
optimization -> optimisation	4
realized -> realised	4
favorite -> favourite	4
realizing -> realising	4
canceled -> cancelled	4
recognizing -> recognising	4
announcer turn relabelled from Guest	4
summarizing -> summarising	4
labor -> labour	4
behaviors -> behaviours	4
colors -> colours	4
organizational -> organisational	4
customized -> customised	3
analog -> analogue	3
optimize -> optimise	3
honor -> honour	2
optimizing -> optimising	2
summarizes -> summarises	2
personalize -> personalise	1
prioritize -> prioritise	1
prioritized -> prioritised	1
organize -> organise	1
rumor -> rumour	1
organized -> organised	1
capitalize -> capitalise	1
centered -> centred	1
organizes -> organises	1
emphasize -> emphasise	1
split name repaired (Schaefer)	1
humor -> humour	1
traveling -> travelling	1
defense -> defence	1
maximize -> maximise	1
Organizations -> Organisations	1
apologize -> apologise	1
digitized -> digitised	1
monetize -> monetise	1
analyzed -> analysed	1
narrator opener re-attributed to Max, words left as transcribed	1
theater -> theatre	1
specialized -> specialised	1
split name repaired (Wadsworth)	1
aluminum -> aluminium	1
specializes -> specialises	1
skeptic -> sceptic	1
customize -> customise	1
personalization -> personalisation	1
optimized -> optimised	1
visualize -> visualise	1
spoken domain -> humanspark.ai	0

What was deliberately left alone

• "Program". A computer program keeps that spelling in UK English; only a broadcast or a scheme is a "programme", and only a reader can tell which was meant.
• "License" and "practice". Both have a noun/verb spelling split in UK English that a rule cannot decide.
• Capitalised words mid-sentence that would otherwise have been respelled: 4 of them, most often part of a name. Left as spoken.
• Retired show and brand names. Spoken by the host and guests. A transcript that removes what was said is not a transcript.
• "Co-pilot". Sometimes Microsoft's product, sometimes the metaphor. A rule would corrupt the honest ones.

The synthetic segments

The opening voiceover on the archive episodes and the closing call to action are read by Max, the show's AI narrator, from a script written in advance. Transcribing them means guessing at words already on file, and the guesses were wrong in places - one opener came back as "an AI that boned him up" where the script says "phoned him up". Those passages carry the authored text instead, and are attributed to Max rather than to the guest: the diariser groups Max with the non-host speaker, which would otherwise put a real person's name on words they never said.

Alastair McDermottI think everybody has an opinion of some kind about the hot mess that is the legal implications of AI, particularly when we're using it to generate content,

but I thought it might be a good idea to speak with an actual lawyer about

Erin AustinThis.

Alastair McDermottAnd I'm delighted to say that Erin Austin is my guest today. Erin, welcome to the AI Powered Thought Leader.

Erin AustinVery happy to be here. Thank you so much for having me.

Alastair McDermottSo everybody has an opinion about all of this kind of mess that is the legal situation with AI, but you are an actual lawyer, so I'm really glad to talk to you about this.

Where are you a lawyer? Where are you based?

Erin AustinYeah, so I'm in the U.S. I live outside of Washington, D.C., but I do serve, you know, federal intellectual property laws, federal laws, so I do serve clients all over the U.S.

Alastair McDermottAwesome. And so your business is called Think Beyond IP, right? So you're the right person to talk to about this.

So this hot mess, as we described it before the show, that is the legal situation with generative AI, where it's taking all of this input in,

whatever we're giving it's got all this training data that we really don't know exactly where all of that came from and whether it was sour ced legally or not.

And then it's generating content for us, and then we don't know exactly when that may or may not be copyrighted, copyrightable, and then we have all of these legal cases.

So where should we start?

Erin AustinWell, when I think about AI and generative AI and how it affects us as experts, as thought leaders, thinking about it from the inputs,

like what's going into the black box that is generative AI, and the outputs, what's coming out of that black box.

And so probably, you know, the most notorious and most talked about are the what's going into the training data, we're talking about those large language models,

there is a massive amount of information that is being used to train those models.

And so, you know, I'm most familiar with chat GBT, I know there are others, but, you know, there is currently a, you know,

some very famous cases, some involving celebrities like Sarah Silverman, you know, using her novel, her book.

And, but one that has risen to the top, and that is very current, and, you know, by the time, you know, the next person watches this, it may have changed.

But as of today, there is a case regarding the New York Times and OpenAI, the company behind chat GBT.

And basically, the training model for chat GBT is approximately like 4% New York Times content. So they have taken all of their content. And that is just a massive amount.

When you think about like, the billions of pieces of information that have to be in a training model, and for 4% of it to be from one entity. That is a huge amount, obviously.

And so, you know, chat GBT or OpenAI, you know, takes the position that it is a transformative use and that it's fair use. Obviously, New York Times has a different perspective.

And importantly, they did a sample prompt.

Granted, the prompt was very directed to get a certain result, but it succeeded in getting an output of an article that was basically word for word, the original content.

And so that is, you know, how we, they start that conversation.

And now the issue is, you know, how much of that was kind of prompt, I wish I remember the term they're using, not prompt fraud, but there's something like that.

So you're, you're kind of gaming the system in order to get a certain result versus something that would happen organ ically when someone would naturally just ask a question to chat GBT and

it would, you know, sort through its billions of pieces of information to come up with a response versus a prompt that is so targeted that it's going to obviously turn up just one source.

And so when we think about that training data, the intellectual property issues, and of course I talk about U. S.

intellectual property issues only, is that an infring ement, a copyright infringement to, you know, take that New York Times information,

taking Silver, Sarah Silverman's book, take scraping the internet, all those things. And is it a copyright infringement?

Is it a breach of the terms, conditions of websites, lots of websites that you can't scrape and they probably are, we don't know exactly how they do it.

And so those are the things that are being litigated right now, you know, from our perspective as users, like we generally,

you know, unless you're making your own data set, and I do believe you can do that with certain people are doing their own.

But generally, we're not, we're not, our concern is on the output end, but the input issue that does concern us is on the prompts.

So we know that, you know, even though they were denying this time last year, that our prompts are also going into that training data. And so what are you putting in your prompts?

Like, if you're working with a client and you have confidential information, are you putting it in the prompts and that becomes part of the data set?

Or your own, you know, proprietary information that you don't want shared with the world?

You know, when I started using ChatGPT first, I'm still just on ChatGPT, I guess I'm a novice, AI-er.

And, you know, it would say, we don't use your training, your prompts, but we know that it was learning. Like, I'd go back and it would know what I did the last time.

So it must be remembering and putting my prompts in there somehow. And so now we should assume, everyone should assume, no matter what AI platform you're using, that your prompts are going,

may make it into the training data, and therefore may be accessible to anyone who has access to that training data. So, and that opens up a whole can of worms when you're

Alastair McDermottUsing information and people's data. Like one thing that I do a lot is I transcribe things like this podcast episode.

And so some of the tools that I use, take this and will actually recognise it.

If I put this into Otter right now, this will recognise my voice and your voice, because your voice is recognised by the system.

Because I had you on my other podcast, the recognised authority, and I tagged your name on there.

And so if I then take that transcript and put parts of that, or the whole thing into ChatGPT, then it is connecting your name with this.

And it could probably figure out who you actually are based on the context of the conversation.

And, you know, it could start to build a thing where it has more about you based on all of this as well, you know?

And that's, that's just, you know, somewhere where we're having an open conversation in public.

The issue is when people are having those conversations, I think in private and in particular, like if you're having calls maybe with customers,

with clients, and you're taking those, or maybe they're really sensitive conversations about people's personal issues. And one side is recording them and transcribing them and

Erin AustinThen putting them into AI.

Alastair McDermottWe don't really know what's going to end up in those systems.

Erin AustinYeah, that is such a good point. Because people are kind of thinking more about the chat GP Ts that we're all taking, but we're using AI all over the place.

Lots of us are using the Anod or Fathom or whatever those note takers are when we are recording our Zoom calls or whatever.

And that is going, that is AI, and it is going into some training data.

And if you are, you know, as a lawyer, and I'm talking to a client, and I'm using one of those, that's very dangerous for me. So I actually don't use them for that reason.

And, you know, people love them. I get it. They make these great notes. They have the recordings, and it's fabulous. But you definitely need a consent issue.

So that your recording, obviously, will say, well, I'm trying to think of all of them require you to consent. I think they all have that pop up.

But that they understand that it's not just like a Zoom call recording, but an AI recording that might go beyond, you know, the two of you.

And so, yeah, we need to be, think about, like, all the places that AI is just showing up. Like, you know, we've been using it for a while without really thinking about it.

But now it's just really part of, like, almost everything that we're doing, a lot of what we're doing when we are interacting online. So we do need to be really careful about that. Yeah.

One thing that I have done in cases where I have something

Alastair McDermottThat's a little bit more sensitive. So, for example, I use Fathom, which is this video recorder that works on Zoom.

And actually, I like the way it actually joins the call as a separate user. And then it puts, Alastair McDermott is recording this call

It has this really obvious notice that this is happening. And I like that about, like, how obvious it is.

But in cases where I have recorded a sales call, and I have used this to record sales calls, it's actually super helpful to record your sales calls and then process them with AI because

you're able to use it to help generate proposals. You're able to use it to improve your sales calls. Like, there's a lot of things that are really useful.

But what I've done is I've taken the transcript and just run it through a text editor and replace the person's name with speaker one or speaker two. And it's a very simple thing.

But that may go some way towards helping to kind of obfusc ate somebody's identity, which if we're putting this stuff in, I think maybe we should be doing that automatically.

We should automatically be hiding the identities of the people who are part of the conversation, unless it's for something, like, obviously public, like a podcast, like we

Erin Austin're doing now.

Alastair McDermottThat's something that I think that maybe should become part of how we just do things.

Erin AustinYeah. Well, I mean, that's interesting.

You know, if, like, let's say in the instance of a legal consultation, and if I have a way of disguising my client, and I have, if there's a way for me to do that, I don't know what it is.

Honestly, you know, I mean, yeah, I can strip it out later, but it's being processed as it's recording. And so that is why I currently am not using it.

I was using them because I thought, oh, this is great. And then I'm like, well, hold on, I can't do this.

And, you know, generally, you know, in the legal industry, we have been, it's been recommended that we never use, whether it's the chat GPTs of the world or the specific legal agreement,

review type of AI programs, like never have any client information in there. It does not matter.

I mean, a lot of those that are specific for the legal industry will say in their terms and conditions that they do not, you know, put any of your inputs into their training data.

And still people, you know, the best practices are just don't do it anyway, because you just don't know.

Alastair McDermottSo, yeah, it's so. And I know, you know, under GDPR here in Europe, like I know that we can't actually use some of these systems.

Like we were not actually able to get access, for example, to the paid version of Anthropics Cloud. So I can use the free version, but not the paid version. And I will sign up to it.

And I'm not able to use the latest test versions of Google's Gemini, which I really want to get access to. So I use a VPN to come in from a US IP address so that I

Erin AustinCan test that.

Alastair McDermottAnd the reason being, I want to test that because it can take up to 1 million tokens, which means it can take about three or four business books worth of data.

And that's really useful to be able to process a large amount of data.

So, yeah, OK, so we're talking about the inputs, the inputs being what we put in in terms of, you know, putting things in like transcriptions with people's names and maybe their personal

data. And then we're and also company information, you know, like, for example, if we go back to the sales call, there might be confidential company information in there as well,

which I would hope that somebody would flag if they see, hey, this is obviously being recorded.

Yeah, maybe maybe we need to think about, you know, put policies in place about what do we do when we join a call and we know there's going to be recorded.

Maybe we say, actually, we're going to talk about some stuff that we don't want to be recorded.

You know, we just need to start to kind of bring these processes into into place and these kind of procedures.

But then the other input is the training data that, you know, we as users didn't have any control over.

That's, you know, Anthropik for their cloud model, Google with all of the data that they have, amounts of data and

Erin AustinOpenAI.

Alastair McDermottAnd I understand that OpenAI have been training their video model on YouTube data.

And I heard on another podcast that Google allowed them to use the YouTube data because Google had already trained their model on YouTube.

And they figured, well, maybe if we're all in the same boat, you know, we won't all get in this much trouble.

Erin AustinWell, that's interesting. I mean, because as we move forward and, you know, clearly we're going into a world where we are going to have more AI.

And so how is, it's not practical considering, again, the volume of information that's in these training models. It's not practical to get a license for all this information.

But I think generally we think that as a public policy position that it's a good thing and that we don't, I don't think there's any undoing what's happened.

Like, I just not going to, it's not going to happen, right? We're not going to make it. It's not going back in the box, no. Right. It's not going back in the box.

So what kind of, you know, kind of mass licensing scheme can we put in place for, you know, AI companies to be able to access all this information and use it?

You know, in the music industry, and I'm not a music lawyer, but we have these compulsory licenses. So I don't have to get permission from the owner of the music.

Like, I can go to these licensing entities, and so long as I pay, you know, them, I can use their music.

And maybe there's something like that that will come about to make sure that, you know, there's some protections in place,

making sure people are compensated somehow for their participation in the training data. But, you know, it's such a massive undertaking. I just don't know exactly how it will shake out.

Alastair McDermottSo if you're talking to somebody and, you know, they're wondering about, you know, what do I need to do at a practical level day to day? What do I need to think about?

What's your advice to them?

Erin AustinYeah, on the input side, make sure, you know, that you have a policy in place about what, how you can and cannot use your generative AI platforms.

So if you're using your own data, you know, certainly this is, you know, as entrepreneurs, we're making our own decisions.

But if we are an employee or an employer, we want to make sure that our employees have clear parameters around what they can use and how and when they can use generative AI.

You know, if it is, you know, writing social media posts, fine.

But if they're, you know, having an analysis of, okay, we're looking at this issue and we want to help, you know, brainstorm it, then, you know, are there some requirements in place there?

Or is there a level of approvals required before we put those things through one of these programs, depending on the level of risk and exposure there is?

When we are using anything with client, that's client information, make sure that it is not a breach of the agreement.

Some people are starting to include language about the use of AI in their services agreements. And so if it says you can't, you know, use it at all, then you can't use it.

And certainly if there's confidential information that you're using, then you putting it through a generative AI platform would be a breach of your confidentiality obligations.

So we need to make sure, be aware of how we are using it on the input side.

Alastair McDermottAnd in terms of, let's say a lot of people are going to be using this for content creation.

And one of the, like, one of the issues I have with this is it's really blurred when it comes to, you know, who owns the output and we've got the U.S. Supreme Court ruling on

Erin AustinThis.

Alastair McDermottActually, can you explain that in layperson's terms for us?

Erin AustinWell, the output side hasn't really changed.

I mean, it's always been that in order for something to be eligible for copyright protection, that it has to be created by a human and has to be original and have a minimum level of

creativity. And so the AI is not considered a human.

And so depending on, and there's no, there's obviously a grey area here because there's everything from something that is, you know, 100% AI created.

And that was the painting that you're talking about. I can't remember the name of it, but where they disclose that this was 100% AI generated painting.

And that was just a very easy, that's not a human and therefore it's not eligible for protection.

You know, the intellectual property laws in the U.S. are grounded on the idea that we want man, you know, include women in there, to be rewarded for the fruit of their intellect.

And that we want to encourage them to be creative and want to give them a certain period of exclusivity to kind of get a return on that investment.

And that is intended to be kind of a right of man, as opposed to a right of machine. You know, famously, there was the photograph taken by a monkey.

Like this was a couple of decades ago and someone tried to, you know, can you copyright that?

No, monkeys don't get, you know, no matter how closely related, they don't get rights under intellectual property law and AI generated works do not get,

even if they are as creative, you know, we're going to assume it's still creative. But if it's not created by a human, then it's not eligible for copyright protection.

So what has been happening is, okay, we're, we're along the continuum of something that's 100% created and something that's, you know, aided.

So let's say you start, you have them generate a outline for you, but then you fill that in or you have them write the first draft for you and you edit it or, you know.

And so whether or not you're going to have copyright protection for that, whether or not it is copyrightable, kind of depends.

And, you know, because it is a case by case, there's no percentage, there's no formula, it will be, you know, you are expected to disclose to the copyright office,

you know, what amount of it is AI generated.

And they will make a determination, you know, whether or not the whole work is protectable, or just the original human created parts are protectable.

And importantly, like, if you know, you try to, you know, don't declare that parts of it are AI generated, and then it becomes,

you know, known later, you know, you can lose their registration, you can lose protection.

And so if you are, you know, creating content, you know, developing your thought leadership using AI, are you creating something that is protectable?

So if you are using it to write a book, and then somebody steals your book, you know, can you, do you have a remedy,

or is it not copyrightable, and therefore in the public domain, and then anyone can use it.

So we want to make sure that if that output is something that we want to have exclusive rights to, and protected under intellectual property laws,

then we want to make sure that we are having sufficient amount of human creativity applied to that work that it would be eligible.

Alastair McDermottSo, and it's subjective.

Erin AustinIt's, it depends, right? It depends.

Alastair McDermottThe best answer. It's because it's always right. Yes, it depends.

Erin AustinYeah.

Alastair McDermottYeah, because like, when I think about this, you know, if you like, if you use AI as a pass through, so what I mean by that is,

if you say, here's some text, fix the spelling and grammar, and you know, it does that. Now, it may, unbeknownst to you, it may actually reword some things.

When you say that, it, you know, it can interpret that as, you know, remove the passive voice or leave the passive voice, and then things start to change a little bit.

And, you know, if you gave it, if you gave it large blocks of text, then maybe that's a small percentage. But if you gave it two sentences, that may be rewriting 40% of what you wrote.

And so, you know, there's all like, and then we get into a situation where like, I use AI with dictation a lot, and I'm giving it my original thinking,

or this is it, this also then becomes a grey area. Like, what is original thinking? Well, it's original thinking, but that's because I was speaking with Christo last night.

And I was thinking about what he said. And then I, and then kind of my brain synthesized that

Erin AustinBased on that.

Alastair McDermottAnd, you know, the 200 books I'd read, and then it kind of mixed it all together. And so is it, is like, is any thinking original?

And yeah, if you're interested in that, go check out the previous episode with Christo, because he's going at it from an artist's perspective and a graphic designer's perspective,

which is very interesting.

Erin AustinFrom a legal perspective, that hasn't changed.

I mean, we've always been influenced by our education, by our experience, by our client work, by, you know, the podcast and the, and where we listen to and the websites that we consume.

And so, you know, but putting our own, you know, kind of spin on things, you know, regardless of how influenced we are,

you know, it's the expression of these ideas and ideas are, of course, are not protectable. But our expression of the ideas is what is protected. And that, and that's true still. Yeah.

Alastair McDermottThis, again, it just seems to be such a mess because we don't know.

And I think that one of the issues is, I think people are just kind of plowing on through it and saying, okay, well, we don't know what's going to happen.

We're just going to keep going forward. That's my take on it.

Erin AustinWell, is there any alternative, right?

I mean, at the end of the day, I hope people are taking advantage of the incredible power of AI and that they're just being mindful about how they're using it.

You know, people obsess about a couple of things that I don't think we need to obsess about.

One is, you know, is AI scraping my website and taking, you know, my work and putting it in there.

And, again, we think about just the massive amount of information that is in a training data set, you know, unlike, you know, the New York Times.

I mean, like, you know, our information is, frankly, the grain of sand in, on the beach, right?

And so, I don't want people to get overly obsessed with, you know, is, are they scraping my stuff and using it? Okay.

and then the other is that now, if it is in there, that somehow people have more access to it than they did before. They have the same access that they ever had with Google.

In fact, personally, I think it's much more diluted, you know, than it was with Google.

If I put in a search term for, you know, AI powered thought leaders in Google, I'm going to get all of your stuff, all of it.

Versus if I put it in a chat GBT, it'll, maybe some of your stuff will be in there, but it'll be like, scrape, it'll be from all of, you know, the data set is what it's pulling from.

And so, I just, I don't want it to stop us from, you know, continuing to, you know, create and not be overly obsessed with, you know, how the AI data sets are using our stuff.

That's, that's just my recommendation there.

Alastair McDermottThat's, that's good advice. I know there's a lot of people kind of afraid of it. And particularly when we see stuff like misinformation and deep fakes and all of this kind of stuff.

Do you have any thoughts on those? Yeah.

Erin AustinWell, that is very important for sure. You know, I'll say like, so this was February, 2023 is the first time I, frankly, I'd ever even heard about chat GBT.

And I was doing a continuing legal education. We have to do these every year to keep our, bar membership.

And I'm like, you know, to me, AI was like the blockchain of, you know, I'm like, I don't, I just, I'll listen to this thing. I don't even know what it is.

And I'm listening and I'm like, what? I can do this. What can I, and so I immediately go to chat GBT and I'm like, oh my God. And so that was it. That was when the love affair started.

But the first time I actually used it, I was looking for an alternative.

I had, you know, a citation for New York law, but I needed the similar citation for a different state, South Carolina. I think it was law.

So I put in like, you know, I have this thing in New York. I need a similar site under South Carolina. So it spits something out for me. It's not a nice, I'm like, oh, great.

And then I'm like, well, let me just read this thing. And it had hallucinated. It completely made up a statute.

And so we need to be really careful about how we're using it for things that require accuracy. You know, so they will, it will hallucinate. It will make errors.

And so from the legal perspective, you know, we are told to treat anything that comes out of, you know, gender of AI as like a junior associate. So you can't just take it and go, thank you.

And, you know, send it off to the client or put it in your brief. You need to review it as if it came from someone who really doesn't know very much.

And, you know, the way I use it is that I will only use it for something if I already know the answer. Like, I know the answer.

I just want you to like, you know, massage it for me or whatever. But never for, you know, advisory services. Like, never I would use it that way.

And so I would, you know, and you can have the reputational hit, you know, even if it isn't an accuracy issue, but it is, you know, on the output side.

There are a couple that will have indemnification for copyright infringement. But most do not. Mostly it's, you know, buyer beware. And you don't know the origin of the output.

Like, there's no, typically, there are no citations. And so what do you have there?

And if you are turning that into a client as original work, you could, you're in breach of that client agreement.

If you're using it for your, you know, your own work, and it is a copyright infringement, you know, you have both legal issues and reputational issues involved.

So we do need to be careful about how we use those outputs.

Alastair McDermottYeah, for sure. And so I think there's two schools of thought on what's going to happen with hallucinations.

I think one is that they're going to be able to figure them out and they're going to go away.

Erin AustinSo don't worry about it too much.

Alastair McDermottAnd then the other school of thought is, actually, no, it's kind of baked into how these systems work, because what they do is they generate stuff.

Erin AustinSo that's the whole point of them. Yes, exactly.

Alastair McDermottSo we're going to, you know, we're going to need to be really careful about what we put in.

Now, the way that I've dealt with this and what I found is that the more accurate, the more data I can feed in, the better the output.

So, for example, I've created custom GPTs where I've uploaded several of my books and all of my blog posts.

And so what I was able to do is I was able to get that to write some and create some content that sounds like me and has my opinion, which is quite interesting.

Oh, I got to talk to you about that.

Erin AustinThat is like my fantasy is to have one of those.

Alastair McDermottLike, is it called a custom or a mini? A custom GPT, yeah.

Erin AustinYeah.

Alastair McDermottThat's what we're doing kind of like at the, you know, at the independent expert solopreneur type business level.

At an enterprise level, they're going to have different solutions where the entire data set can be trained just on their enterprise data.

But those are going to be custom solutions that are going to be very expensive.

But for people like you and me and, you know, who have small businesses, small teams, we don't want to, like, we don't have the need or the resources for something like one of those

enterprise systems. But being able to have your own data in there is really useful. And, yeah, so I found that very useful for me in, again, in writing first drafts of things in particular.

And I created, I actually created one, which is based on my books, but my book and my work, my workbook about podcasts. And I used it while I was driving. I used it in the conversation mode.

I just had my headset and I was talking to ChatGPT on my own custom GPT. And I got it to plan this podcast, including, you know, all of the, we actually created some of the graphics.

Erin AustinWow. I noticed you had AI graphics at the beginning.

Alastair McDermottYeah. So everything on this show is AI generated. And I did that because I kind of felt like I have permission to do that on this show. And I wanted to experiment a bit with it.

But, yeah, it's, you know, again, then, like, then we get into that whole situation with the copyright, you know.

Do I lose copyright to the whole thing because I've got the AI generated section right at the start? And that's subjective, right?

Erin AustinWell, if we just talk about this episode, let's just talk about this episode. So everything that we are creating here as humans, you know, that is all protectable.

Is your opening that if it's all AI generated? No, but if you were to register this episode with the copyright office, you would say, like, it's a podcast episode.

And there's, like, a box where you would say, like, the part material that's excluded. And you'd say kind of the images at the beginning. But your protection would apply to everything else.

So it's a whole thing out. Yeah. Interesting. Interesting. Yeah.

Alastair McDermottAnd so I think that a lot of people who are creating content are using the kind of the implied copyright protection just on creation rather than actually contacting the copyright office.

Erin AustinRight. Yeah, I mean, think about how much content, I mean, how much you make every day, how much I make every day. We don't register most of that stuff.

And it's only if, you know, something happens and there's a contest of some sort that it would ever come up, honestly, right? So go ahead.

If you have your AI generated content, do your thing with it, you know, put your copyright notice on there if you want to.

But if you do want to go that extra step of registration, then you really do need to parse it so that you are being honest.

Your application, you are making statements that under penalty of perjury that what you're disclosing is correct.

But if you're just, you know, it's another social media post or another blog post that you're not registering, just, you know, do your thing.

Alastair McDermottIs there anything that I haven't asked you about when it comes to the legal implications of using AI? Anything that we should cover?

Erin AustinYeah, I mean, just generally, I think going back to the issue of if you're using it to create something that you are being paid for or that you want to be paid for,

that is not the works that you want to use AI generated content for because it may not be protectable.

So again, you know, having the copyright protection, the remedies that you have under intellectual property law, that requires it to be, you know, created by a human and original.

And so, you know, make your outlines with it, you know, do some brainstorming with it.

But the content itself has to be copyrightable if you want to make, build a business from it, you know, like if it's something that you want to license to a third party,

can it clear due diligence? If you're using AI to create those materials that would be of value to a third party, they have to be original to you and protectable. And this is all subjective.

Alastair McDermottIs that right?

Erin AustinWell, it depends on, I mean, it is, there is a continuum, right?

But I do believe that most of us have a pretty good idea of how much original we are adding to an outline or to a first draft. Yeah.

Alastair McDermottI'm thinking specifically of the case that my use case is where I'm dictating something and then I'll have an AI rewrite it. In fact, I might have it rewrite it several times.

And so it looks very different, but the original thinking

Erin AustinCame from me.

Alastair McDermottAnd I think this one is kind of, it's right in the middle. That's where I think it's subjective, right?

Erin AustinYes. Yes. Because we don't know what happened when you put it in and what came out. Like maybe it took what you said and went, oh, I see this article here.

That's, you know, and then they kick, it kicks that back out. So that we don't know.

So what I would say is after you have the output, the, what you would do with it between the output and publication is where the originality in creative has to come. So, yeah. Yeah. Yeah.

Alastair McDermottAnd I think for people who are using this day to day, I mean, the, it's still, it's still so new that the output that we're getting right now from tools like ChatGPT 4 is, it's okay.

It's not brilliant.

you know, in terms of, you know, I say that and like the wonderful things I could do, you know, it's kind of getting very jaded now in this, like it's incredible in what I can do.

But when we look at the text output that it gives us, it doesn't look like it's quite written by a human yet.

And it certainly, like, it's not at the point where I would want to use it directly. I would always want to edit it. So that edit step is really important.

And then the amount of input that we gave it is also a big factor here. If we want it to be copyrightable, that's, that's what I'm taking from what you said.

Erin AustinYeah. I mean, and just from purely from the thought leadership branding perspective, you know, you know, it needs to be in your voice.

Like, so whatever that final product is, you want it to be in your voice. And probably, I mean, certainly not today, you're not going to get any output that's going to do that.

And so you're going to put that layer on it at minimum.

Alastair McDermottAnd that's where the custom GPT that I was talking to you a bit earlier about was, is quite useful because if it has,

you know, a hundred thousand words of your writing in your voice already, it's able to use that to figure out how to do more like that. So it can actually get it, get it quite a lot closer.

Okay, so input, careful about what we're putting in, both from a privacy perspective, confidentiality, and, you know,

having some policies around that and having like a clear dialogue with clients.

If we're putting client information in there, then in terms of output, how we're actually using it and the level of protection that we have over that output depends really on what we did in

terms of input.

Erin AustinYes, exactly. Yeah. But just definitely be careful. I mean, if it's, you know, your social media post of the day, you know, eh.

But if it's something that's going into a book or going into a course, you know. Yeah. Right.

Alastair McDermottSo maybe we have different levels of kind of different levels of attention or I'm not quite sure how to articulate that, but we need to judge each on its own merit.

Erin AustinYeah, like, I mean, I like to look at it if it's like a moneymaker, you know, a direct moneymaker, like, versus if it's,

you know, just some marketing material, however green it is, you know, even if it's not a moneymaker, but it's something that,

you know, like a, you know, manifesto of some sort that you give away for free.

You know, things like that you want to, you know, the more valuable it is long term, the more attention you pay to that output.

Alastair McDermottThe other thing you were talking about was policies. And I think it's really important to put in place some kind of policies.

Have you talked to people about that, about what they're doing with, you know, internal policies for these things?

Erin AustinYeah, I mean, only, you know, I've heard a lot of it has been on the legal side, like how law firms are doing it.

Honestly, I'm just starting to have conversations with how kind of the consultant uses it.

But that is, you know, kind of next up, I've been like doing some research, it was very timely conversation today, because I have been looking at that people are asking lots of questions

about that. And so I am working on a resource that will help people kind of navigate like the things that they need to be thinking about, and how to put those policies in place.

Alastair McDermottYeah, and they can find you at think beyond IP.com. So I'm sure Erin will have, she has a great newsletter

Erin AustinThere.

Alastair McDermottOkay, so I think this has been fascinating, a little bit scary, of course.

Erin AustinOh, don't be scared.

Alastair McDermottIt's still, it's still, like, I feel like it's still messy.

Erin AustinOh, it is for sure. I mean, there's so many, yeah, there's so many questions to be answered. Like, there are more questions than answers, of course. Yeah. And so, yeah.

Alastair McDermottAnd that's going to settle over time, right? As more cases meet the court, as people figure out, more people start using these tools, those will eventually get

Erin AustinIroned out.

Alastair McDermottYes, exactly.

Erin AustinYeah. And it may be, it depends. I mean, think about fair use. Fair use has been around forever, and it's still complete, you know, one court's fair use is another court's infring ement.

So, I don't know if we're going to get a better answer than that, but, you know, we learn to work with it. Okay.

Alastair McDermottSo, my takeaway is, you know, we started off and this thing was a bit of a mess. Now it's a little bit of a more organised mess.

Erin AustinMy work here is done.

Alastair McDermottErin Austin, thank you for joining me today on the AI Power ed Thought Leader. It's nice to chat with you again.

You can, if you're listening to this or watching this, you can find me and Erin. We're both on LinkedIn and very active over there. We both have newsletters.

I have a free guide to getting started with AI. If you're just getting started, you can go to learnai.guide. You can find that on there. And definitely sign up for Erin's email list.

And she has a great newsletter there that I recommend if you are thinking about using intellectual property in your

Erin AustinBusiness.

Alastair McDermottWhich you should be. Erin, thank you so much for being with us today.

MaxThanks for listening to Something About AI. If you got something out of this one, share it with someone who would get something out of it too. That is genuinely how the show finds people.

There is more like this coming, so do subscribe wherever you are listening. And there is a pile of free resources over at humanspark.ai. Go and have a look.