Latest / Inside DuckDuckGo / Duck Tales: 32% of AI users tell AI things they've kept from other people. Our new study on AI's privacy problem (Ep.42)
Transcript
- 0:00SPEAKER_00: Hey, everyone.
- 0:01SPEAKER_00: Welcome to DuckTales,
- 0:02SPEAKER_00: where we go behind the scenes at DuckDuckGo and discuss the stories,
- 0:06SPEAKER_00: the technology, and the people that help build the privacy tools for everyone.
- 0:11SPEAKER_00: In each episode,
- 0:11SPEAKER_00: you will hear from employees about our vision,
- 0:14SPEAKER_00: product updates, engineering, marketing, but also our approach to AI.
- 0:20SPEAKER_00: Today, you're going to hear about some brand new research,
- 0:22SPEAKER_00: very fresh, that we ran about what Americans actually think about privacy and AI.
- 0:28SPEAKER_00: And I'm here today with someone who happens to have the exact same haircut as me.
- 0:33SPEAKER_00: And that means luscious locks, of course.
- 0:35SPEAKER_00: Zach, do you want to introduce yourself?
- 0:38SPEAKER_01: Yeah.
- 0:38SPEAKER_01: Hey, everybody.
- 0:39SPEAKER_01: I'm Zach Pappas, SVP of User Insights.
- 0:42SPEAKER_01: I've been at DuckDuckGo for 14 wonderful years,
- 0:46SPEAKER_01: and I'll be talking to you about this study today.
- 0:48SPEAKER_00: Wow, 14 years.
- 0:49SPEAKER_00: Yeah, we're very lucky to have you here.
- 0:52SPEAKER_00: You've been working in Insights for a long time.
- 0:56SPEAKER_00: You're one of the very first DuckDuckGo employees.
- 0:58SPEAKER_00: Is this correct?
- 1:00SPEAKER_01: Yeah, I think it was technically the second employee back when we were just
- 1:04SPEAKER_01: a few people around the Paoli office here in Pennsylvania.
- 1:07SPEAKER_01: So it's been a long 14 years, but I think it's been...
- 1:11SPEAKER_01: Really exciting.
- 1:12SPEAKER_01: And things change so frequently that it doesn't really feel like we're doing
- 1:16SPEAKER_01: the same insights or kind of same research over and over again.
- 1:19SPEAKER_00: Well, that's pretty good.
- 1:21SPEAKER_00: But it's also useful because it means you also have about as long as a view
- 1:25SPEAKER_00: as anyone on how people feel and think about privacy,
- 1:28SPEAKER_00: what they want protected, what they say versus what they do.
- 1:30SPEAKER_00: So I think this is going to be an interesting conversation.
- 1:34SPEAKER_00: And yeah, by the way, you've introduced yourself.
- 1:36SPEAKER_00: I've been talking.
- 1:37SPEAKER_00: I have been rude.
- 1:38SPEAKER_00: I haven't introduced myself.
- 1:39SPEAKER_00: My name is Mario.
- 1:40SPEAKER_00: I'm on the marketing team.
- 1:42SPEAKER_00: I've been here for almost four years now.
- 1:45SPEAKER_00: A fun fact about myself is that I like to tell people I once came in fourth
- 1:51SPEAKER_00: in Eurovision Song Contest.
- 1:52SPEAKER_00: For those who don't know,
- 1:53SPEAKER_00: it's like it's big music kind of song festival and competition in Europe.
- 1:59SPEAKER_00: So yeah, I like to say I once came in fourth, but kind of, so not exactly.
- 2:03SPEAKER_00: And I'll just leave it at that.
- 2:05SPEAKER_00: What about you, Zach?
- 2:06SPEAKER_00: Do you have any fun facts you want to share?
- 2:08SPEAKER_00: Or obnoxious maps?
- 2:10SPEAKER_01: Nothing that interesting or standout.
- 2:14SPEAKER_01: I guess,
- 2:15SPEAKER_01: like most people,
- 2:16SPEAKER_01: when I have the time,
- 2:17SPEAKER_01: I will go out to my garage, fire up the forge and do a bit of blacksmithing.
- 2:22SPEAKER_00: Wow.
- 2:23SPEAKER_01: So pretty standard stuff.
- 2:25SPEAKER_00: Wow.
- 2:25SPEAKER_00: Okay.
- 2:26SPEAKER_00: That's quite cool.
- 2:27SPEAKER_00: I mean...
- 2:28SPEAKER_00: This is DuckTales, not ZackTales.
- 2:31SPEAKER_00: If it was ZackTales, I would ask you a lot more questions.
- 2:33SPEAKER_00: Maybe put in a request.
- 2:34SPEAKER_00: Next episode.
- 2:35SPEAKER_00: Yeah, next time.
- 2:36SPEAKER_00: Okay, let's jump in.
- 2:38SPEAKER_00: So we as a company,
- 2:41SPEAKER_00: and you particularly,
- 2:42SPEAKER_00: since you've been here for so long,
- 2:43SPEAKER_00: we've been studying how people think about privacy for close to,
- 2:46SPEAKER_00: I don't know, in this case, like 15 years or more.
- 2:49SPEAKER_00: You know, what's new is doing that specifically for AI.
- 2:54SPEAKER_00: In June,
- 2:55SPEAKER_00: our User Insights team surveyed 1,944 U.S. adults,
- 3:01SPEAKER_00: all quota balanced to census on age, gender, region,
- 3:03SPEAKER_00: et cetera,
- 3:05SPEAKER_00: about their AI privacy concerns,
- 3:07SPEAKER_00: what they know, what they don't know, what they actually do about it.
- 3:11SPEAKER_00: You'll find the methodology on the show notes if you're interested.
- 3:13SPEAKER_00: And this is what we want to talk about today.
- 3:16SPEAKER_00: So, Zach, let's start with the juicy stuff.
- 3:19SPEAKER_00: To you, what was the most surprising thing in this study?
- 3:25SPEAKER_01: Uh...
- 3:27SPEAKER_01: I think given the fact that AI has been so pervasive and around for so long,
- 3:31SPEAKER_01: most surprising thing was that I think 43% of everyone knew none of six basic facts
- 3:38SPEAKER_01: about how AI chats are handled.
- 3:40SPEAKER_01: So that includes a third of active AI users who can recall or kind
- 3:47SPEAKER_01: of signal what basic data privacy and data handling practices that the AI services
- 3:53SPEAKER_01: that they're using are actually adhered to.
- 3:57SPEAKER_00: Can you tell us a little more like what facts you're talking about?
- 4:00SPEAKER_01: Yeah, I actually wrote them down so I can get them correct for you.
- 4:04SPEAKER_01: So the facts we asked people about were AI chats can be subpoenaed by courts
- 4:10SPEAKER_01: and law enforcement.
- 4:11SPEAKER_01: AI chats can be made public in a lawsuit.
- 4:14SPEAKER_01: Employers can access worker enterprise AI accounts.
- 4:17SPEAKER_01: Settings to delete or stop storage vary by provider.
- 4:21SPEAKER_01: Free accounts often have weaker privacy than paid accounts.
- 4:25SPEAKER_01: And humans can review your chats.
- 4:28SPEAKER_01: Some of these maybe you would expect most people who haven't dug too far into,
- 4:32SPEAKER_01: you know, AI or privacy policies that maybe wouldn't think of those things first.
- 4:36SPEAKER_01: But certainly human employees can review your chats.
- 4:39SPEAKER_01: We've seen instances of this in the past with other companies and data collection
- 4:43SPEAKER_01: when there's a breach or when there's a mishandling of data.
- 4:46SPEAKER_01: Those are some pretty basic things that I would have assumed people at least,
- 4:50SPEAKER_01: you know, had some suspicions about.
- 4:52SPEAKER_01: But it's surprising that they did.
- 4:54SPEAKER_00: Yeah, this is a little, yeah,
- 4:58SPEAKER_00: it was a little unexpected,
- 5:00SPEAKER_00: the level of, at least to me, you know, listening to this.
- 5:04SPEAKER_00: I was looking a little bit here at the details.
- 5:07SPEAKER_00: Like I saw,
- 5:07SPEAKER_00: for example,
- 5:09SPEAKER_00: that the AI chats can be subpoenaed by the government,
- 5:11SPEAKER_00: for example,
- 5:11SPEAKER_00: which means they could then be accessible in public because there were a lot of news
- 5:14SPEAKER_00: about this, about some cases about AI companies.
- 5:17SPEAKER_00: It was about only 25% of people knew this.
- 5:21SPEAKER_00: Why do you think people don't know these facts, given it has been so much on the news?
- 5:25SPEAKER_01: I know this isn't in the
- 5:26SPEAKER_00: study, but just curious about
- 5:27SPEAKER_01: what you mentioned.
- 5:28SPEAKER_01: Yeah, I mean, it's definitely a different technology.
- 5:30SPEAKER_01: I think it's been a lot more opaque the way that it's been integrated and with
- 5:35SPEAKER_01: the speed that it's been rolled out and the kind of the breadth
- 5:38SPEAKER_01: that it's been rolled out to most products that people are using.
- 5:40SPEAKER_01: They get a different flavor and a different version of it.
- 5:42SPEAKER_01: AI everywhere that they look with incomplete data practices,
- 5:46SPEAKER_01: meaning at the executive level,
- 5:49SPEAKER_01: most companies haven't really figured out how to deploy AI safely.
- 5:53SPEAKER_01: So what they're getting is,
- 5:54SPEAKER_01: let's say,
- 5:54SPEAKER_01: a rushed experience,
- 5:55SPEAKER_01: but something that probably wasn't built in the best interests of data privacy
- 5:59SPEAKER_01: or kind of long term sustainability in that regard.
- 6:03SPEAKER_01: So here, yeah,
- 6:04SPEAKER_01: it's maybe a combination of it being kind of a pervasive,
- 6:09SPEAKER_01: like broad applied technology, as well as how fast it has risen.
- 6:14SPEAKER_01: Maybe my own personal experience is there isn't that much useful information for
- 6:20SPEAKER_01: a mainstream person,
- 6:21SPEAKER_01: maybe who's not a developer or a coder,
- 6:23SPEAKER_01: who's going to be really in the weeds of the technology.
- 6:26SPEAKER_01: It's kind of hard to find some of the better,
- 6:29SPEAKER_01: more mainstream or general approaches than how to use it safely.
- 6:33SPEAKER_00: Yeah, yeah, I can see that.
- 6:34SPEAKER_00: Thanks for that.
- 6:36SPEAKER_00: So, yeah, we know what's the most surprising thing about the study for you.
- 6:39SPEAKER_00: But asking the bigger question, like, are people worried about privacy in AI, in the AI context?
- 6:45SPEAKER_01: Yeah, absolutely.
- 6:47SPEAKER_01: 75% of people are worried about privacy when using AI.
- 6:50SPEAKER_01: And that has been consistent for years as we've been running these studies.
- 6:54SPEAKER_01: And we've seen this with external research as well.
- 6:57SPEAKER_01: Invasion of privacy, just the discomfort and uneasiness of using Yeah, and...
- 7:22SPEAKER_00: Yeah.
- 7:22SPEAKER_00: And I remember some of these studies that are published like outside of our own.
- 7:26SPEAKER_00: And it's always like main concerns about AI.
- 7:28SPEAKER_00: And there's usually one that's like, you know, economic impact or job losses.
- 7:32SPEAKER_00: And then privacy is usually hanging there on the top.
- 7:35SPEAKER_00: So I definitely.
- 7:36SPEAKER_01: Yeah, absolutely.
- 7:37SPEAKER_01: And, you know, I think.
- 7:40SPEAKER_01: That gets lumped in with a lot of things,
- 7:42SPEAKER_01: particularly the thing that I think most people do end up worrying about is
- 7:46SPEAKER_01: that it's going to have some impact on their financial situation down the line,
- 7:50SPEAKER_01: whether that's your personal details being involved in a breach,
- 7:54SPEAKER_01: credit card details,
- 7:56SPEAKER_01: somebody getting into or unfreezing your credit account and trying to open
- 8:00SPEAKER_01: a new line of credit.
- 8:01SPEAKER_01: Like a lot of these things at the end of the day,
- 8:02SPEAKER_01: I think get back to the comfort and stability that people have
- 8:05SPEAKER_01: with either their finances or their way of life.
- 8:08SPEAKER_01: Is it going to jeopardize?
- 8:09SPEAKER_01: Yeah, and I also remember other studies that you guys have run where
- 8:38SPEAKER_00: It isn't like there's a one single thing about privacy that people are worried about.
- 8:42SPEAKER_00: There's a broad concern about privacy.
- 8:44SPEAKER_00: And for some people, it's generic on like, I don't like the surveillance.
- 8:49SPEAKER_00: It's an invasion of privacy.
- 8:50SPEAKER_00: It can be a little more amorphous.
- 8:52SPEAKER_00: For others, it's a bit more concrete.
- 8:54SPEAKER_00: But the main thing is that there's a broad general concern on this topic.
- 9:00SPEAKER_00: And so there's a big concern about privacy.
- 9:03SPEAKER_00: 75% of people worry about it when using AI.
- 9:05SPEAKER_00: What are they doing about it?
- 9:09SPEAKER_01: It's a tough story to tell.
- 9:12SPEAKER_01: Not tough because we don't have the data or that we don't understand it.
- 9:15SPEAKER_01: It's just like kind of hard to piece through.
- 9:17SPEAKER_01: So about 40% have taken no action at all to protect themselves.
- 9:22SPEAKER_01: But what that means is 60% have and about half of them of
- 9:25SPEAKER_01: that 60% have like Just either avoided using AI or maybe toned down a question
- 9:32SPEAKER_01: because of privacy concerns.
- 9:33SPEAKER_01: And so that kind of nets out to about one in three of the U.S.
- 9:37SPEAKER_01: adults that we polled at least have held back from using AI in cases
- 9:40SPEAKER_01: where they've had a privacy concern.
- 9:43SPEAKER_01: Given the volume,
- 9:44SPEAKER_01: the number of people using AI,
- 9:45SPEAKER_01: that is a lot of potential interactions every day that people are either withholding
- 9:49SPEAKER_01: or kind of transforming into a way that they feel more safe in using it.
- 9:54SPEAKER_01: Others are actually seeking out the privacy options that exist.
- 9:58SPEAKER_01: There are some things like temporary chats,
- 10:01SPEAKER_01: switching between multiple accounts or services so that you're not contaminating,
- 10:06SPEAKER_01: so to speak, one AI with some context that you might not want it to have.
- 10:11SPEAKER_01: But where there aren't strong defaults, we can expect that to be an uphill battle.
- 10:15SPEAKER_01: As we saw this with search privacy and browser privacy is that people tend
- 10:19SPEAKER_01: to default to whatever the first thing is that they see in the service
- 10:23SPEAKER_01: that they're using.
- 10:24SPEAKER_01: So it's no surprise that the number isn't higher because it's not easy for people
- 10:28SPEAKER_01: to find,
- 10:28SPEAKER_01: even though they are demonstrating a lot of pent-up demand and action
- 10:33SPEAKER_00: on it.
- 10:34SPEAKER_00: Yeah, this is interesting.
- 10:35SPEAKER_00: I remember once being in a qualitative study and I was just talking to people
- 10:42SPEAKER_00: and they're talking about their concerns.
- 10:44SPEAKER_00: And I remember there were a few people that they were saying they would,
- 10:48SPEAKER_00: for example, they would often...
- 10:50SPEAKER_00: When they're asking something to AI, they'll even change what they're writing.
- 10:55SPEAKER_00: So they will not write what they want to ask.
- 10:57SPEAKER_00: They will use some code words, for example.
- 10:59SPEAKER_00: So people are kind of editing themselves.
- 11:04SPEAKER_00: Yeah, for sure.
- 11:06SPEAKER_01: You know,
- 11:06SPEAKER_01: another big part of this that we didn't study specifically here,
- 11:09SPEAKER_01: but that I can imagine plays a large part in the psychology
- 11:12SPEAKER_01: behind this is not having a very clear policy,
- 11:14SPEAKER_01: even on the work-administered AI or physical organization that you're a part of.
- 11:20SPEAKER_01: They've given you access to a company tool.
- 11:23SPEAKER_01: Usually there's some concern there about what the admin of that
- 11:26SPEAKER_00: tool can see,
- 11:28SPEAKER_00: how the company is collecting all
- 11:29SPEAKER_00: of that data, even from their own employees.
- 11:31SPEAKER_01: So it kind of goes back to the idea that they're not given the right instruction or
- 11:36SPEAKER_01: the right success set up to get them started with AI in a way that's actually going
- 11:42SPEAKER_01: to help them long term and abate this issue of them having a concern
- 11:45SPEAKER_01: about their privacy the whole time that they're using it.
- 11:47SPEAKER_00: Yeah, of course, of course.
- 11:50SPEAKER_00: So, yeah, people have these concerns.
- 11:53SPEAKER_00: Do they trust the AI company?
- 11:55SPEAKER_01: No, I knew you were going to ask that.
- 11:58SPEAKER_01: No, it's 14% of respondents say that they mostly or completely trust AI companies.
- 12:04SPEAKER_01: I think it's 39% have no trust at all.
- 12:08SPEAKER_01: And that's been, you know, so one of the more consistent findings.
- 12:10SPEAKER_01: I think it's a lot of different findings,
- 12:11SPEAKER_01: especially I think it gets worse as companies continue
- 12:14SPEAKER_01: to demonstrate some pretty poor handling of the rollout of AI,
- 12:19SPEAKER_01: both from a privacy perspective and also just a UX perspective.
- 12:21SPEAKER_01: I think it contributes a lot to the mistrust that people have in the tool
- 12:24SPEAKER_01: when they can't really understand why it's being integrated
- 12:28SPEAKER_01: into their product experience.
- 12:34SPEAKER_01: AI hasn't really been integrated that much, at least in a salient way.
- 12:38SPEAKER_01: You can look at a stat, this U.S. government trust in AI was even worse.
- 12:43SPEAKER_01: It was 48% have no trust at all in the U.S.
- 12:47SPEAKER_01: government to mostly or completely, or to protect their data.
- 12:51SPEAKER_00: Yeah, and I can imagine it's not just U.S. government specifically.
- 12:54SPEAKER_00: I can see it being generalized even in other places, even though this is just a U.S. study.
- 12:59SPEAKER_01: Yeah, I mean,
- 13:01SPEAKER_01: maybe my other personal take on this is there really aren't a whole lot
- 13:04SPEAKER_01: of good actors or, you know, role models in the space.
- 13:08SPEAKER_01: I hope DuckDuckGo can really stand for that and show people how to do it well.
- 13:12SPEAKER_01: But if you think about people who are doing,
- 13:15SPEAKER_01: you know,
- 13:15SPEAKER_01: good AI rollouts,
- 13:17SPEAKER_01: particularly the way that they're handling,
- 13:19SPEAKER_01: you know, user data or other types of data, there aren't that many great examples.
- 13:23SPEAKER_01: I can't, you know, think of any right now.
- 13:26SPEAKER_00: Yeah.
- 13:27SPEAKER_00: Okay.
- 13:27SPEAKER_00: So, We have three quarters of people
- 13:31SPEAKER_00: that are, you know, of people that are concerned about privacy when using AI.
- 13:36SPEAKER_00: We have about two thirds of the US using AI in some way, shape or form.
- 13:41SPEAKER_00: So how do you kind of,
- 13:43SPEAKER_00: you know,
- 13:44SPEAKER_00: how do you think about these two,
- 13:47SPEAKER_00: you know,
- 13:47SPEAKER_00: similarly contradicting, you know, stats if folks don't trust AI with their data?
- 13:53SPEAKER_00: Like, how do you explain this?
- 13:55SPEAKER_01: It's a good question.
- 13:56SPEAKER_00: That's the classic privacy paradox.
- 13:58SPEAKER_01: But as with the past,
- 14:01SPEAKER_01: I think it's kind of a load of BS in the sense that people still use things
- 14:05SPEAKER_01: that they don't trust, especially when they don't have another choice.
- 14:08SPEAKER_01: So a good example might be like your cable company.
- 14:11SPEAKER_01: I don't think most people have a really high affection for their cable company,
- 14:17SPEAKER_01: yet they are forced to use it.
- 14:19SPEAKER_01: They use it a mess.
- 14:22SPEAKER_01: So oftentimes there aren't other options.
- 14:23SPEAKER_01: And if DuckDuckGo is an example,
- 14:25SPEAKER_01: people would rather use something that doesn't feel like they're paying for it
- 14:28SPEAKER_01: with their data.
- 14:29SPEAKER_01: And I think it was the stat in our study that said 58% of AI users agree
- 14:34SPEAKER_01: that like privacy keeps them from doing more with AI than they currently do.
- 14:37SPEAKER_01: And it goes back to that stat of the one third of Americans kind
- 14:41SPEAKER_01: of like preventing themselves from putting something into AI in one of those cases.
- 14:45SPEAKER_01: Like that's a lot of this pent up demand and uneasiness
- 14:48SPEAKER_01: that shows you people are still using the tool,
- 14:50SPEAKER_01: but they could be using it more happily or in a more delightful way
- 14:54SPEAKER_01: that doesn't come with that trade off to their concern.
- 14:57SPEAKER_00: Yeah, yeah.
- 14:58SPEAKER_00: Totally.
- 14:59SPEAKER_00: I can see that.
- 14:59SPEAKER_00: And so, okay, clear.
- 15:01SPEAKER_00: There's concerns.
- 15:04SPEAKER_00: You know, there's a privacy paradox.
- 15:06SPEAKER_00: Is there anything,
- 15:06SPEAKER_00: like when you look at this broad data called concerns,
- 15:09SPEAKER_00: trust, is there any pattern when you look at specific segments?
- 15:12SPEAKER_00: Is there, you know, are Are these concerns common across different segments?
- 15:17SPEAKER_01: Yeah.
- 15:18SPEAKER_01: It's funny.
- 15:19SPEAKER_01: This was a question that was asked a long time ago when I first started
- 15:21SPEAKER_01: because we used to look at the same thing with regards to search privacy
- 15:25SPEAKER_01: or browser privacy.
- 15:26SPEAKER_01: Is privacy more pernicious or Yeah.
- 15:28SPEAKER_01: Yeah.
- 15:46SPEAKER_01: You know, apply some some basic rigor to that.
- 15:49SPEAKER_01: I think you could say,
- 15:50SPEAKER_01: you know,
- 15:51SPEAKER_01: people who are less technically inclined and the older folks or people
- 15:54SPEAKER_01: who are less educated might have more of an aversion to it in certain cases or
- 15:59SPEAKER_01: for certain applications.
- 16:01SPEAKER_01: But broadly,
- 16:01SPEAKER_01: it is something that stretches across all walks of life that we that we at least
- 16:05SPEAKER_01: that we've studied.
- 16:08SPEAKER_00: Thank you.
- 16:08SPEAKER_00: Thank you for that.
- 16:09SPEAKER_00: So, OK, moving from this this this area.
- 16:16SPEAKER_00: When you go through the data in the study,
- 16:17SPEAKER_00: was there any other fact that you think would surprise people if they would hear
- 16:24SPEAKER_00: about it?
- 16:25SPEAKER_00: Anything that came out of the study, any insight that would surprise anyone reading through it?
- 16:30SPEAKER_01: Yes, and we are going to put something up on our spread privacy blog about this.
- 16:35SPEAKER_01: The stat that I thought was most surprising was 32%
- 16:38SPEAKER_01: of all AI users have shared something with AI that they have not shared with
- 16:42SPEAKER_01: a family member, friend, or like a doctor.
- 16:45SPEAKER_01: So that just kind of goes to tell you that there is this large contingent of people
- 16:49SPEAKER_01: who are kind of giving away more to AI than they would even to
- 16:53SPEAKER_01: the most trusted confidants that they have.
- 16:56SPEAKER_00: So in a way, they are using AI as a confidant more than humans sometimes.
- 17:01SPEAKER_00: So, yeah, it's not surprising.
- 17:04SPEAKER_00: I mean, you know, just, you know, just talking to people and seeing the news and I can see that.
- 17:10SPEAKER_00: But is this something again, I'm going to ask you the segment question.
- 17:13SPEAKER_00: Is this something that is more prevalent on certain groups and less so on others?
- 17:18SPEAKER_01: Well,
- 17:19SPEAKER_01: earlier you were asking about privacy concerns or privacy concerns are prevalent
- 17:22SPEAKER_01: across pretty much every psychographic demographic.
- 17:25SPEAKER_01: And like I said, something I think everybody feels on usage and the way that they're using it.
- 17:30SPEAKER_01: The interesting thing that we found was that parents are actually confiding.
- 17:33SPEAKER_01: Mm-hmm.
- 17:55SPEAKER_01: You know, lucrative information from them.
- 17:58SPEAKER_01: Partly, I think because,
- 17:59SPEAKER_01: you know,
- 17:59SPEAKER_01: there's a lot of certain life stages where that level
- 18:02SPEAKER_01: of expertise is both like expensive or inaccessible.
- 18:07SPEAKER_01: So parents probably are asking a lot of questions that might be
- 18:09SPEAKER_01: a little bit embarrassing or a little bit too difficult to reach out to somebody
- 18:13SPEAKER_01: about maybe requires a lot of context.
- 18:15SPEAKER_01: And so here you see this behavior with parents kind of confiding more in AI compared
- 18:20SPEAKER_01: to non-parents.
- 18:20SPEAKER_01: And I do expect that if we look in other areas, we would see that as well.
- 18:25SPEAKER_00: Yeah, and that is a very interesting stat,
- 18:27SPEAKER_00: and it reminds me once when I was talking with a few,
- 18:30SPEAKER_00: let's say,
- 18:32SPEAKER_00: AI users who were parents from another study,
- 18:34SPEAKER_00: and I remember them talking about the tension of sharing things with AI
- 18:38SPEAKER_00: because it's useful in many ways, but at the same time, there's also concerns.
- 18:41SPEAKER_00: So that's just interesting to have this one.
- 18:44SPEAKER_00: Thank you for this one.
- 18:45SPEAKER_00: So you mentioned also before AI training, like AI chats being used to train AI.
- 18:51SPEAKER_00: So you You mentioned that in general people didn't know about
- 19:01SPEAKER_00: any of these facts
- 19:02SPEAKER_00: about AI, but specifically about this one, about training.
- 19:05SPEAKER_00: How many people know about it?
- 19:07SPEAKER_00: How do people feel about it?
- 19:09SPEAKER_00: Because this is something that's talked about a lot.
- 19:11SPEAKER_01: Yeah,
- 19:11SPEAKER_01: so the earlier stat was 40% of people can't name even one of the six basic stats
- 19:17SPEAKER_01: about how the data is handled.
- 19:18SPEAKER_01: But if you look specifically at the criteria of whether or not people are aware
- 19:26SPEAKER_01: that AI services can train based on your data,
- 19:30SPEAKER_01: 53% of people didn't know that,
- 19:31SPEAKER_01: or they weren't sure that their conversations were used for AI training.
- 19:35SPEAKER_01: And when they learn about it, of that 58%,
- 19:39SPEAKER_01: It's, you know, about half of them or more than half of them are really uncomfortable with it.
- 19:42SPEAKER_01: And I think 30% are very uncomfortable.
- 19:45SPEAKER_01: So it's something that I think as people do realize that
- 19:48SPEAKER_01: or find out how it's being applied and learn that these,
- 19:51SPEAKER_01: you know,
- 19:52SPEAKER_01: that if you take that group of people who had no idea about any
- 19:55SPEAKER_01: of these six concerns,
- 19:56SPEAKER_01: the more exposure that they get to it,
- 19:58SPEAKER_01: the more that they learn about it, the more concern that they're likely to develop.
- 20:01SPEAKER_00: Yeah, and my awareness of the issue, and it's so important.
- 20:06SPEAKER_00: Okay, that's interesting and also good to understand.
- 20:11SPEAKER_00: So we've been talking a lot about the study,
- 20:13SPEAKER_00: and I think we can almost start finishing the conversation here on the study.
- 20:18SPEAKER_00: But I'm just curious,
- 20:18SPEAKER_00: especially with your long experience studying privacy and technology,
- 20:24SPEAKER_00: how is studying privacy?
- 20:25SPEAKER_00: Is AI privacy different from studying privacy in other technology or studying
- 20:30SPEAKER_00: the consumers' and users' opinion about it?
- 20:34SPEAKER_00: What's different from what you've been studying for many, many years?
- 20:39SPEAKER_01: It's such a great question, Mario.
- 20:42SPEAKER_01: There's a lot that's different.
- 20:43SPEAKER_01: There's also a lot that's similar.
- 20:44SPEAKER_01: I guess one of the hard things is there's always a background rate
- 20:47SPEAKER_01: of people's misunderstanding between where certain products and others begin.
- 20:51SPEAKER_01: So search to browse was historically one of the kind of boundaries
- 20:54SPEAKER_01: that I think people had a hard time with.
- 20:56SPEAKER_01: Where does Google Chrome And Google search start would be kind of something
- 21:00SPEAKER_01: that we would try to get our heads around.
- 21:02SPEAKER_01: Here at the simplest level, people just give AI more.
- 21:05SPEAKER_01: It is a chat box with a lot of text entry.
- 21:09SPEAKER_01: It makes it very easy to just give it kind of more information.
- 21:12SPEAKER_01: Right.
- 21:12SPEAKER_01: Right.
- 21:12SPEAKER_01: Right.
- 21:27SPEAKER_01: Right.
- 21:42SPEAKER_01: That data means it's very opaque.
- 21:44SPEAKER_01: It's not completely unknown how the AI companies themselves are handling or storing that data.
- 21:50SPEAKER_01: And then you come back to that risk.
- 21:52SPEAKER_01: If there is a breach,
- 21:53SPEAKER_01: if there is some bad actor or an employee who has access to this,
- 21:57SPEAKER_01: what they have access to is a completely different level of detail and context
- 22:03SPEAKER_01: than prior.
- 22:04SPEAKER_01: So in researching this,
- 22:05SPEAKER_01: something that we try to take into account is, one,
- 22:07SPEAKER_01: this kind of background rate of people who You know,
- 22:10SPEAKER_01: share either a deep concern,
- 22:12SPEAKER_01: share a deep concern as well as a like low understanding of exactly what
- 22:17SPEAKER_01: the concern should be about.
- 22:19SPEAKER_01: And like that,
- 22:20SPEAKER_01: sussing that out a little bit more,
- 22:21SPEAKER_01: I think is something that we're going to have to get a lot better at over time
- 22:24SPEAKER_01: because the products are going to get more complicated and in my opinion,
- 22:27SPEAKER_01: more homogenous.
- 22:28SPEAKER_01: Almost every app that I use today has some kind of a chat box in it.
- 22:33SPEAKER_01: They all look very similar.
- 22:34SPEAKER_01: They all act differently.
- 22:35SPEAKER_01: Somewhat similar.
- 22:36SPEAKER_01: And at the end of the day,
- 22:37SPEAKER_01: I think what people are really trying to get back to is a solid understanding
- 22:40SPEAKER_01: that their lives are getting better
- 22:42SPEAKER_01: and they're not trading off something extreme like their privacy or their well-being
- 22:48SPEAKER_01: as a result of that.
- 22:49SPEAKER_00: Yeah, the black walks.
- 22:52SPEAKER_00: This reminds me of,
- 22:53SPEAKER_00: I remember once watching Chanel,
- 22:57SPEAKER_00: one of our researchers,
- 22:58SPEAKER_00: and she was talking to someone and this person was saying that with AI,
- 23:06SPEAKER_00: They were saying,
- 23:06SPEAKER_00: if someone sees what I'm typing on a search box or on my browser,
- 23:11SPEAKER_00: they'll know what I do.
- 23:13SPEAKER_00: But if they know what my AI chats are, they know how I think.
- 23:18SPEAKER_00: So there is this feeling of a lot more intimacy and sensitivity and exposing
- 23:23SPEAKER_00: or putting your brain on a paper,
- 23:25SPEAKER_00: like it was even an expression I've heard someone use,
- 23:27SPEAKER_01: that I think it connects well to what you said.
- 23:30SPEAKER_01: Yeah, exactly that.
- 23:33SPEAKER_00: Yeah.
- 23:34SPEAKER_00: Well, cool.
- 23:35SPEAKER_00: Well, Zach, I think this ends the study discussion.
- 23:38SPEAKER_00: You know,
- 23:38SPEAKER_00: it's good you give us this context,
- 23:39SPEAKER_00: but, you know, this is a little, it's a little, it's a little dire.
- 23:44SPEAKER_00: Like, where does this leave,
- 23:45SPEAKER_00: you know,
- 23:45SPEAKER_00: internet users and people,
- 23:48SPEAKER_00: you know,
- 23:48SPEAKER_00: just want to use it to,
- 23:49SPEAKER_00: you know, improve their lives and, you know, and yeah, essentially.
- 23:55SPEAKER_01: Do you have any words for us
- 23:57SPEAKER_00: on that?
- 23:57SPEAKER_01: Yeah, hopefully these people will end up finding out about DuckDuckGo
- 24:03SPEAKER_01: because we've made IA optional and private by default.
- 24:06SPEAKER_01: So this is something that,
- 24:07SPEAKER_01: you know,
- 24:08SPEAKER_01: because this is our bread and butter,
- 24:09SPEAKER_01: we've tried to understand it and really make a product that serves the And I think
- 24:13SPEAKER_01: a lot of companies are going to have to meet users in the middle.
- 24:16SPEAKER_01: Otherwise, people,
- 24:17SPEAKER_01: as they normally do,
- 24:20SPEAKER_01: will vote with their dollars or kind of walk out of these products in search
- 24:24SPEAKER_01: of something better.
- 24:40SPEAKER_00: Yeah, yeah.
- 24:41SPEAKER_00: Okay, that's good to end on this, on a trust and hope.
- 24:45SPEAKER_00: And I'm glad we're,
- 24:46SPEAKER_00: you know,
- 24:47SPEAKER_00: as a company that we're trying to help in making things better for everyone.
- 24:52SPEAKER_00: So, okay, so my,
- 24:55SPEAKER_00: you know,
- 24:56SPEAKER_00: I think like if I could summarize this,
- 24:58SPEAKER_00: people are worried and check me and check if I'm saying something wrong.
- 25:01SPEAKER_00: People are worried about AI and privacy.
- 25:05SPEAKER_00: They're worried,
- 25:06SPEAKER_00: but they still use AI,
- 25:08SPEAKER_00: even often as a confidant that they will say things that they don't say to others.
- 25:13SPEAKER_00: When you do inform people about the specifics of how they are at risk,
- 25:18SPEAKER_00: they don't like it at all.
- 25:22SPEAKER_00: People that are already concerned about privacy,
- 25:24SPEAKER_00: many are taking action,
- 25:26SPEAKER_00: for example, by not asking certain questions or using AI less.
- 25:32SPEAKER_00: And this is kind of like what I'm taking from it.
- 25:36SPEAKER_00: And again,
- 25:36SPEAKER_00: it's good to know that,
- 25:38SPEAKER_00: you know,
- 25:38SPEAKER_00: companies like DuckDuckGo,
- 25:39SPEAKER_00: but others as well, are building tools to solve this issue.
- 25:45SPEAKER_00: And this is a better place to be than,
- 25:47SPEAKER_00: you know,
- 25:47SPEAKER_00: nobody cares that
- 25:49SPEAKER_01: there's people acting on it and that
- 25:50SPEAKER_00: people are also taking action
- 25:51SPEAKER_00: for it.
- 25:54SPEAKER_00: And yeah, so Zach, I...
- 25:56SPEAKER_00: Thank you for this.
- 25:57SPEAKER_00: Thank you for this overview.
- 25:58SPEAKER_00: But I don't want to end on this note.
- 26:02SPEAKER_00: So we're going to try to do something we have never done on DuckTales.
- 26:06SPEAKER_00: Are you willing?
- 26:08SPEAKER_01: Yeah, this is my first DuckTales.
- 26:10SPEAKER_01: I don't know what you're prepared for, so let's do it.
- 26:12SPEAKER_01: I'm excited.
- 26:13SPEAKER_00: Okay, very good.
- 26:14SPEAKER_01: So you've heard about Subway Takes, like popular online series.
- 26:18SPEAKER_01: Like Hot Takes?
- 26:19SPEAKER_00: Yeah, Subway Takes, Hot Takes.
- 26:21SPEAKER_00: Yeah, or Hot Takes.
- 26:23SPEAKER_00: That works well.
- 26:24SPEAKER_00: So, yeah.
- 26:25SPEAKER_00: So we're going to adapt this format a little bit.
- 26:28SPEAKER_00: I'm going to give you a topic.
- 26:30SPEAKER_00: You're going to give me your Hot Take.
- 26:32SPEAKER_00: And then I'll tell you if I agree with it or I disagree with it.
- 26:36SPEAKER_00: Okay?
- 26:36SPEAKER_01: The most important part.
- 26:38SPEAKER_00: What's honestly the most relevant part?
- 26:40SPEAKER_00: We're going to call this Zach Takes.
- 26:42SPEAKER_00: It's a new segment.
- 26:43SPEAKER_00: Very good.
- 26:45SPEAKER_00: Ready?
- 26:46SPEAKER_00: Yes.
- 26:46SPEAKER_00: Okay.
- 26:46SPEAKER_00: So what's the most overrated thing in AI right now?
- 26:53SPEAKER_01: Frontier models.
- 26:55SPEAKER_01: Do I have to explain these?
- 26:56SPEAKER_01: Should I explain them?
- 26:56SPEAKER_00: Just lightly.
- 26:57SPEAKER_00: Just lightly.
- 26:58SPEAKER_01: Frontier.
- 26:59SPEAKER_01: I just think if what you were using yesterday was...
- 27:03SPEAKER_01: Now, all of a sudden,
- 27:04SPEAKER_01: invalidated by something that was released today, then we kind of lost the point.
- 27:08SPEAKER_01: And I understand this is a little bit of a loaded question,
- 27:10SPEAKER_01: and I don't get enough time to respond to it.
- 27:12SPEAKER_01: But I'm going to say frontier models.
- 27:13SPEAKER_01: I don't think there is a big of a deal.
- 27:14SPEAKER_00: 50% agree.
- 27:16SPEAKER_00: 50% agree.
- 27:17SPEAKER_00: That's where I'll leave it.
- 27:18SPEAKER_00: What's the most underrated thing in AI right now?
- 27:23SPEAKER_01: Same.
- 27:23SPEAKER_01: This is where I was talking about task completion.
- 27:25SPEAKER_01: I think that is the thing that we should be talking about and whether
- 27:27SPEAKER_01: or not people are actually getting something better than they were getting before.
- 27:31SPEAKER_01: Can they order pizza faster or better, more consistently?
- 27:35SPEAKER_01: Can they converse with their family about things that they couldn't do before?
- 27:39SPEAKER_01: I think we kind of lost sight of the basics
- 27:41SPEAKER_01: of building good products is satisfying task completion.
- 27:45SPEAKER_00: I'm going to give you another 50% degree on this one.
- 27:49SPEAKER_00: I would counterpoint with, I think, giving people control of if, how, and when they use AI.
- 27:55SPEAKER_00: For me, that's like an underrated feature,
- 27:58SPEAKER_00: let's put it like that, that I think we could all benefit.
- 28:02SPEAKER_00: So in five years,
- 28:03SPEAKER_00: another one, in five years, do you think private AI is going to be its own category?
- 28:07SPEAKER_00: It's going to be a checkbox inside every big AI product?
- 28:12SPEAKER_01: Oh, no.
- 28:14SPEAKER_01: I would say both.
- 28:15SPEAKER_01: I think even, I mean, now it is a table stakes feature.
- 28:18SPEAKER_01: I think that people are expecting,
- 28:21SPEAKER_01: even though most companies aren't delivering it or delivering it transparently.
- 28:24SPEAKER_01: So I would say that it's both.
- 28:26SPEAKER_01: And I do think before long,
- 28:28SPEAKER_01: we'll see a lot more vertical integrated AI where somebody is going to try to make
- 28:32SPEAKER_01: a play either for privacy or some level of trust, which is the big differentiator.
- 28:37SPEAKER_00: Okay.
- 28:38SPEAKER_00: I can't not say 100% agree on that one.
- 28:40SPEAKER_00: So final two, what's an everyday object that most deserves a privacy policy?
- 28:48SPEAKER_00: Um...
- 28:52SPEAKER_01: The easy answer would have been like mirror or something.
- 28:55SPEAKER_01: But I'm going to say a hotel key card because, and this is just a personal thing.
- 29:00SPEAKER_01: Every time I have a hotel key card, it is, I don't know if I should throw it away.
- 29:03SPEAKER_01: Do I return it to the desk?
- 29:05SPEAKER_01: Should I shred it?
- 29:07SPEAKER_01: I've heard that there's nothing on there.
- 29:08SPEAKER_01: It's just, I don't know.
- 29:09SPEAKER_01: But at this point, I would just like to know.
- 29:11SPEAKER_01: But information is important.
- 29:13SPEAKER_01: Yeah, there's maybe a little disclaimer on the back of the card
- 29:15SPEAKER_01: that could just tell me exactly what's on there.
- 29:17SPEAKER_01: I'd feel a little bit better about how I can do it.
- 29:20SPEAKER_00: Very good.
- 29:20SPEAKER_00: Finally, last one.
- 29:22SPEAKER_00: Last one of probably the last episode of Zach Dates ever.
- 29:25SPEAKER_00: What is the single most correct way to spell Zach?
- 29:33SPEAKER_01: There's no way to know.
- 29:33SPEAKER_01: That knowledge was lost centuries ago.
- 29:38SPEAKER_00: Zach, I'm going to say.
- 29:39SPEAKER_01: I did in high school.
- 29:39SPEAKER_01: I don't know if this person is going to be watching this,
- 29:43SPEAKER_01: but I did in high school try to make a play for ZAQ as the way to spell Zach.
- 29:47SPEAKER_01: 100% agree.
- 29:48SPEAKER_00: That's it.
- 29:50SPEAKER_00: We found the answer.
- 29:51SPEAKER_00: We found the answer.
- 29:52SPEAKER_00: No way I need it.
- 29:53SPEAKER_01: I haven't seen anybody pull it off.
- 29:54SPEAKER_01: I don't think I am the type of person to pull it off,
- 29:56SPEAKER_01: but if I would have started in the early 2000s,
- 29:59SPEAKER_01: I think by this point, it would be fashionable.
- 30:01SPEAKER_00: Very good.
- 30:02SPEAKER_00: Well, nicely done, Zach.
- 30:03SPEAKER_00: Anything else you want to add before we wrap this?
- 30:05SPEAKER_01: Nothing for me.
- 30:06SPEAKER_01: Thank you so much, Mara.
- 30:07SPEAKER_01: This is great.
- 30:07SPEAKER_00: Very good.
- 30:08SPEAKER_00: So, Zach, thank you so much for joining DuckTales.
- 30:10SPEAKER_00: And to everyone listening,
- 30:11SPEAKER_00: if you've got feedback on the show,
- 30:13SPEAKER_00: episode idea,
- 30:15SPEAKER_00: you want to complain about the host,
- 30:16SPEAKER_00: want to know more about this study,
- 30:17SPEAKER_00: for example, write to us at podcast at duckduckgo.com.
- 30:21SPEAKER_00: That's podcast at duckduckgo.com.
- 30:25SPEAKER_00: We read all of them.
- 30:26SPEAKER_00: We will listen to them if it's an audio recording.
- 30:28SPEAKER_00: We will see the video in case you're sending us a video.
- 30:31SPEAKER_00: We're multimodal.
- 30:33SPEAKER_00: So, thanks, Zach, and see you around.
- 30:36SPEAKER_00: Thanks.