AI deepfakes: can your reputation survive them?

About the episode

It looks like your CEO. It sounds like your CEO. It isn't your CEO. And millions of people just watched it.

"You can easily just pay a subscription of $20 and create a fake of a brand," says UNSW Business School’s Associate Professor Eric Lim. That's the reality of the deepfake threat, and it's already here, from cloned voices to a viral fake ABC news report.

A business can spend decades building its reputation, only for an AI clone to hijack it in seconds. Deepfakes are now so ingrained in our online world that it's hard to tell what's real and what's a convincing scam.

You can't stop AI fakes from being made, and A/Prof. Lim says detection tools will always be one step behind. So the strongest defence has nothing to do with technology: "you start with the individual, because that's the only defence that you can control."

The Business Of is brought to you by UNSW Business School, produced with Deadset Studios.

Want to know more? 

Want to follow the latest research and news from UNSW Business School and AGSM @ UNSW Business School? Subscribe to BusinessThink and follow UNSW Business School and AGSM on LinkedIn. Or listen to more episodes of The Business Of and subscribe to our UNSW Business School and AGSM channels on YouTube.

Edited transcript

A/Prof. Eric Lim  00:02
The people are going to fall for deepfakes over and over again, no matter how advanced your technical defences are, or how much education you provide to your people in the company. I guarantee that they will happen again. It's a never-ending war. So, where do you start?

Dr Juliet Bourke  00:21
In the age of deepfakes, anyone can clone a voice, fake a face, and produce content to sound exactly like someone you trust, even me. That was an AI version of my voice, and if it sounded convincing, you've already understood the problem. Deepfakes are no longer just a Hollywood issue; they're cheap, fast, and accessible. Recently, a fake news report went viral featuring finance expert Alan Kohler and Commonwealth Bank CEO Matt Comyn. It looked like the ABC, sounded like the ABC, but it was a scam designed to steal your trust and your money, so we're in an age where you can't believe your own eyes and ears. How does a business prove it's the real thing? I'm Dr Juliet Bourke, Adjunct Professor at UNSW School of Management and Governance, and this is The Business Of, a podcast by the UNSW Business School. A/Prof. Eric Lim, founder of the UNSW Crypto Clinic, says he has the answer. Eric, let's start at the beginning. What exactly is a deepfake?

A/Prof. Eric Lim  01:43
Deepfakes essentially use AI technology, or, actually, the innovative technologies we see these days, to create fake realities. It could be a video, an image that has been doctored, and so on and so forth, with some intentions behind it, whether for entertainment or simply because people have malicious intent in creating those deepfakes and disseminating them.

Dr Juliet Bourke  02:14
If I think about deepfakes, I think about them in the digital world. But then I think about paintings, and there's always been master fakers. How is it any different today?

A/Prof. Eric Lim  02:25
So, you're right, conceptually they are the same, right? But people are concerned with the velocity at which these fakes evolve, and also, they are concerned with the ease with which these fakes appear and are created. I think these are the two main factors that people are worried about, rather than the existence of fakes, which has always been there. There is easy access to all these tools. You can easily just pay a subscription of $20 and create a fake brand of a company doing something that is damaging to itself, or create a persona that is related to a company, so that the ease with which these fakes are created is probably causing a lot of disturbance in the minds of businesses.

Dr Juliet Bourke  03:16
So it's just the pervasiveness of the technologies that's the threat, but what are the actors behind it wanting to do?

A/Prof. Eric Lim  03:24
My assumption is that they are predominantly divided into two intentions, right? One intention is that it could be entertaining, which means, basically, I think this is funny. I want to put it out on social media and make people laugh. The other dimension is malicious, in the sense that they could profit from creating deepfakes. Recently, there was a story of ABC where Alan Kohler and Matt Comyn, CBA CEO, were having a fight in the studio, and when Matt was called out for having a secret investment platform, or something, so I believe that that was intentionally malicious, in a sense that people who created that deepfake was probably trying to benefit from it by driving traffic to that fake investment platform that Matt supposedly was investing in.

Dr Juliet Bourke  04:30
So there's this moment we're in with AI-generated clutter, and I guess the question is: how do brands demonstrate their credibility and authenticity? When there's the chance that you're actually looking at a deepfake?

A/Prof. Eric Lim  04:48
This is a good question, isn't it? In a sense, that's how you defend against deepfakes that are trying to damage a company's reputation. Even a personal reputation, if you have one that you want to protect. Ultimately, I think it comes down to your track record, your reputation, in a sense, how resilient your reputation is. You know, if you demonstrate, if, as an individual, you demonstrate that you are a trustworthy person, that means you go about your life, and when you are, you're not perfect. No one is perfect, right? You carry out your actions, and if you get something wrong, you correct yourself, you omit your mistakes, and so on and so forth. People are more willing to trust you, but if you're the opposite, when you're caught misbehaving or making a mistake, you just kind of double down and say, 'I wasn't wrong,' people are less likely to believe that. So, if I could bring in a little personal example, my grandfather is 106 years old. Actually, I think he's an example. He leads his life well, his children, and his countless grandchildren, and I admire him. And if he did something wrong, I'm sure that his reputation will survive, because he's simply a great man.

Dr Juliet Bourke  06:22
Do you want to talk about real-world examples of people facing threats to their reputations and overcoming them, or even coming out looking good?

A/Prof. Eric Lim  06:36
I'm not sure if you remember the case of Johnson and Johnson, where you know they pull all their Tylenol off the shelves, because there were a few cases where the Tylenol has been tampered with, and people who have bought them died from that poisoning, so it was a deliberate sabotage event, and Johnson & Johnson was seen as decisive, they were seen as being more concerned with the lives and the safety of their customers than profit, in the sense that they're willing to take a hit by pulling everything off the shelves at a moment's notice. So I think that event gave them a huge, formidable, resilient reputation.

Dr Juliet Bourke  07:22
And how does that help us know, as business leaders, how to deal with a deepfake threat?

A/Prof. Eric Lim  07:29
So, if I want to relate that to deepfakes, I would say that if you are a company today in the environment where deepfakes are so prevalent, you might want to act in good faith, right? You want to be a moral character in society, and to be judged by your actions, to garner a good track record of all the things that you do, and the things that you say, and that will give you resilience. And then you know that will protect you against unforeseen circumstances in the future. It could be from deepfakes or other things.

Dr Juliet Bourke  08:05
If I think back to deepfakes, is the link that when a deepfake happens, a consumer looks at that one instance in the context of longevity and says, "Actually, that's unusual, that can't be right, that's right, it can't be that Matt Comyn threw a glass of water in someone's face, that's just not the right kind of character. Is that what we rely upon now?

A/Prof. Eric Lim  08:30
Yes, not simply because of the fact that you know it's good to protect your own reputation or understand the importance of protecting a company's reputation or brand in that sense, but simply because of the fact that everything starts from the individual. If you want a company to have resilient brands and a resilient reputation, you need good people, people with good character.

Dr Juliet Bourke  08:56
So, I hear what you're saying about following individuals, and we follow their reputation. What happens? I see all these Hollywood influences, Hollywood actors and actresses, and then there's a threat to that reputation, perhaps through a deepfake. How do I make sense of that?

A/Prof. Eric Lim  09:13
Yeah, Hollywood is a very interesting area, right? In the sense that the stars have their own personal brands. In fact, they are like a company in their own right, so, in that sense, they have all the incentives to protect their own reputation. Sadly, it seems like there's a lot of glitz in Hollywood but not much wisdom. You can see a lot of Hollywood stars just self-destruct right by doing very silly things over and over again to the point where you know they become a toxic asset, and no one wants to touch them. All the Hollywood studios just avoid them, whereas you get stars that have staying power. You get stars like Keanu Reeves, who I think has a great reputation, as he seems down-to-earth. Seems to be trying too hard to fake it, but again, I would say that that has to be coupled with your own personal understanding of what authenticity means.

Dr Juliet Bourke  10:13
So the Italian Prime Minister recently had a deepfake made about her, and of course, there were some people who bought into that, but her strategy was actually to publicise that it was a deepfake and to put it up on X and say this is what happens to people, this is not me, so to kind of own it and challenge it in her own right. What do you think about that as a strategy?

A/Prof. Eric Lim  10:36
Well, it really depends, right? I mean, if she already had a stellar reputation or has a track record, then by coming out to own it, and to say that I have nothing to hide or go check it out, I think that's brilliant, right? But if you already have a sordid reputation and you put it out there, it's just kind of like drawing more attention to yourself, and people are more likely to believe that than not. I mean, that's kind of how I see it logically.

Dr Juliet Bourke  11:09
Let's get back to business, and I'm thinking about business owners who have a brand and want to protect it.  Are there layers of defence that they have to put in place to make sure they're not attacked?

A/Prof. Eric Lim  11:24
I never believed that the defence against deepfakes would be a single solution; it can't be right. It has to be multifaceted, so in a way we could have technological solutions, or at least technological defence, where you can have AI calculating the probability that you know something is fake, and so on and so forth, and you can have blockchain technologies that check for the provenance of the information, in the sense that if something that is said by a politician has to be verified, we can all trace back to the provenance of that information, and then, of course, you can have, like I said, a great track record, so it's multifaceted, right? You can have technological solutions, social solutions, and individual defences against deepfakes, so it can't just be one particular solution. We can all learn from the examples we have had in protecting against cybersecurity threats. You see technological solutions, and so on and so forth, but it still happens, no matter how advanced your technical defences are, or how much education you provide to your people in the company. I guarantee that they will happen again. So, same thing, deepfakes, it's a never-ending war. So, where do you start? That's why my proposal is to start with the individual, because that's the only defence you can control.

Dr Juliet Bourke  12:50
Eric, your research is looking at solutions based on decentralised ID, which sounds like a brand verification badge. Can you explain how that works?

A/Prof. Eric Lim  13:00
Well, the concept of decentralised identity is that you want everyone, right, every entity, so it could be an individual, it could be a company, it could be a government institution to have a ID on a blockchain that they can control, and in a sense that you can issue statements, you can issue content, and kind of like tie that content back to your ID on the blockchain, such that everybody can verify the provenance of that content and whether that content has been tampered with or not. So that's the technological solution, that's one technological solution that that we have explored, but that is mainly for the verification of the provenance of that information, so it's not like a centralised platform like Facebook or X, where let's say if you go against certain regulations or certain company policy, they can delete the content, or they can, you know, shadow ban your content, and so on and so forth. So, imagine if you have a presence in those platforms, and you have made a name for yourself, you have created this brand for yourself, and suddenly you're shut down, your whole livelihood is gone. So that's the danger of a centralised platform: with a decentralised one, no one can really shut you down.

Dr Juliet Bourke  14:36
So Edelman's Trust Barometer is an annual index that looks at who we trust. We used to trust institutions and leaders, but it seems we're trusting influencers today. These are people whom we don't know, but they've got a following. Does that make us more susceptible to deepfakes?

A/Prof. Eric Lim  14:57
I don't think so. I think. The turning to influencers is actually a response to the failure of our institutions, our politicians. We just feel that we can't trust them anymore. I think the reason why deepfakes are so prevalent and so threatening is because we have lost that character, that ability to say that there are wrong things in the world, and there are right things in the world, and there are authentic things in the world where you feel in your bones that it's right, this is correct, this is true, and this is not right, this is wrong, right? I mean, if we bring that back, we can alleviate the detrimental effects of these deepfakes. People are less likely to fall for it. There are, like I said, digital solutions, you know, AI fighting AI, in the sense that AI calculates the probability that this Butler digital artefact is actually fake or not fake, but again, they were never going to be able to calculate the probability; they can't tell for certain, for certainty anyway. But, as I said, that's just one aspect of how you can defend yourself against deepfakes, but the most important aspect of it, I feel, comes from the individual.

Dr Juliet Bourke  16:21
And what do you think the future looks like then, in terms of deepfakes? What's the next iteration of what we're in now,  if we're in a problematic context?

A/Prof. Eric Lim  16:32
Try to draw lessons from, as I said earlier, about cybersecurity. I mean, we've been talking about cybersecurity for years; it's not a new phenomenon, right? We keep saying that. Okay, we're gonna have a new detection tool. If you apply that to the same context of deep fix, it's the same narrative. We say that there's a deep fix, there's a threat from deepfakes, and then we have all these technological defences against it, or we can actually educate people against, you know, looking out for telltale signs, where this is a deep fix. We can educate people better. Now, if you think about it, that is exactly the same narrative as we have for the cybersecurity defence. We say, well, we can have technological solutions, training, and so on. But why do cybersecurity breaches still happen? Are they not good enough? Obviously, they're not good enough. That's why it happens. So, it's the same thing with deepfakes. People are going to fall for deepfakes over and over again. I can almost guarantee it. That's how I see it.

Dr Juliet Bourke  17:33
I love that you've got this philosophical take on things: the answer is not digital. The answer is to go back and reflect on ourselves. How have we lost the ability to determine authenticity from inauthenticity? That's your basic thesis.

A/Prof. Eric Lim  17:49
That's my basic thesis, in a sense, that when you're authentic, I'm not saying that you're going to get it right, but at least it increases your probability of getting it right, that's what I'm saying.

Dr Juliet Bourke  18:04
So, what do you think a business leader needs to do to protect themselves from deepfakes, which, as I understand it, are going to keep on happening?

A/Prof. Eric Lim  18:13
Yeah, deepfakes are going to keep happening, but I think the best line of defence is to instil a deep, authentic foundation in individuals, whether they're the company leader or people within the company. Once you have that firm foundation, you're going to realise that you can detect deepfakes better, not 100%, but you can do it better.

Dr Juliet Bourke  18:44
That's Associate Professor Eric Lim from UNSW Business School. If you want to understand what happens when AI sets the price you pay, listen to our episode with UNSW Associate Professor Sam Kirshner.

A/Prof. Sam Kirshner 18:57
Why we feel very unfair if you look at the kind of dynamic pricing example from the airlines is that it is just so opaque in terms of what they are actually using to sell me this flight at this price, so what is it that's in my cookies or my browsing behaviour or just the country or city that I'm in, that is determining this price?

Dr Juliet Bourke  19:24
You'll find the link in the episode description. The Business Of is brought to you by the UNSW Business School, produced with Deadset Studios.

To stay up-to-date with our latest podcasts as well as the latest insights and thought leadership from the Business School, subscribe to BusinessThink.