How Does Apple’s Siri Compete In an Agentic World?

[00:00:03] Welcome to Tech News Briefing. It's Friday, October 16th. I'm Danny Lewis for The Wall Street Journal. We've officially entered the era of the personal AI agent, and it feels like every other day, another tech giant launches their entry into the bot race. But one name is noticeably missing from the hype. Apple.

[00:00:24] Today, we're bringing you a conversation with cybersecurity reporter Bob McMillan, personal tech columnist Nicole Nguyen, and Apple reporter Rolf Winkler to break down why Apple is so far behind on AI and what it would take for them to catch up. And before we get started, a quick note. News Corp, owner of The Wall Street Journal, has a commercial agreement to supply news through Apple services. On to the show.

[00:00:48] Hi, I'm Bob McMillan. I'm a cybersecurity reporter here at The Wall Street Journal, and I'm joined today by my colleagues, Nicole Nguyen and Rolf Winkler. Hi, Bob. Hello. Good to have you guys here. You know, I was watching this show, Ted Lasso. Have you ever watched that show? I haven't, actually. You haven't watched Ted Lasso? I haven't watched Ted Lasso. You're an Apple— I'm the last— You're a person who covers Apple, and you haven't watched Ted Lasso? Yeah. One of my favorite characters, Keely Jones, there's this scene in the show where she picks up her phone and she talks to Siri.

[00:01:17] And she asks it like a—sort of like a chat GBT kind of question, like, you know, will I be a good mom kind of thing? And I was like, it's so weird to see somebody use Siri for something that's not just like directions or send an email. You know what I mean? Like, do you— And that's how you know it's an Apple show. You know it's an Apple show because they have to get the product placement. Yeah. But I don't use Siri. And do you use Siri?

[00:01:41] I use Siri about a bazillion times a day to set timers and almost nothing else. I think in testing this new revamped upgraded Siri, when Siri AI rolled out last month, I had to force myself to use Siri. But I often just reverted to using chat GBT or Claude to answer the question. Yeah. I guess—what do I use it for?

[00:02:03] When I'm driving my kids around, I want to play a song on Spotify with my voice as opposed to having to look down at it or call a phone number. And it works most of the time, but it's about the same for those use cases as the old Siri. If you wanted to do a chatbot experience and ask it questions and have a conversation the way you do with ChatGPT or Claude, it can do that now. And it can give you competent answers.

[00:02:33] But the goalposts have already moved. Well, this is the theme, I think. If there's a theme to our episode today, it's Apple, the AI company. And to me, Siri is the premier example of Apple as an AI product, even though they didn't invent it. And it's not widely used in the way that other chatbots are used. But I want to first start, Nicole, we've got to talk about some of the stuff from the week.

[00:02:58] And you had this great story about Apple's new—the iHome? Was it the name of it? It's called HomeView. And it looks like a giant Apple Watch or a baby iMac G4. Do you remember from the early aughts, that domed, white-domed iMac with the articulating display? It's like a little baby mini version of that. It has a 7-inch square display. And it's supposed to replace the iPad as a smart home hub.

[00:03:28] Wait, so is that the end of the iPad or just as a hub? Just as a smart home hub. Okay. But it's also a way for Apple to expand its reach of Siri AI into your living room or kitchen or bedroom. How many people actually use any of the HomeKit, you know, smart home products from Apple? Apple is a distant third in smart home. That's true. Although, like, nobody in Apple land can still mean, like, millions of people.

[00:03:57] So I think in, like, the relative scheme of things, Amazon and Google are years ahead of Apple in terms of adoption, the number of smart home devices that are compatible with their respective hubs and their respective ecosystems. And Apple has sort of thought of the home platform as, you know, ancillary to the iPhone. So what's the key to succeeding in the home hub market?

[00:04:26] Like, why has Apple not been as successful there? It hasn't come out with its own smart home devices. So, you know, Google has the Nest camera and both Google and Amazon have had their own smart displays, which have been successful for them in this, like, post-LLM world. Yeah. Because now the response is – can be long and can be detailed. And so you kind of want to see that response on a screen. And so, you know, Apple just didn't have any presence there.

[00:04:54] You could stick an iPad on the wall and use that, I suppose. But there's something different about having a dedicated, always plugged in, always listening device that lives on your countertop, and it serves the purpose of helping you set timers in the kitchen mostly. Well, isn't that the problem, though, that it's not actually going to help you solve all that many problems? Because no third-party apps, right? So there are some third-party apps.

[00:05:23] It's the same audio apps that work in CarPlay. So if you have, like, a podcast app or an audiobook app that also works in CarPlay, it'll work. This new release really represents how tightly controlled Apple likes to, you know, keep its ecosystem. I think even a lot of Apple fans reacting to this device launch were just struggling to understand why would I want this. Why would I want this? Yeah. And they're just like, wait a minute.

[00:05:50] What if I just took my iPhone, I put it on a stand, and I have that in my kitchen? I have all my apps. I can still use FaceTime. I can use Siri. I can use all these things I want to do. I can control my smart home from there. But I – so why am I going to get this separate device? And I told Rolf. And the answer is? And I told Rolf it's because you need a hub without getting into it because the smart home is such a mess. And if you want to be a systems admin in your own life, like, if you want to be your personal IT person in your house, then set up a smart home.

[00:06:19] But it's such a mess that, like, not a lot of people do it. But you still need a hub, and that's why you'd need this home view rather than just putting your iPhone on the wall. Do I need a hub? I mean, I have ring cameras. I have – I had Nest thermostats that we just ripped out for our new heating system. Casa smart switches. A garage door opener that has an app. I mean, all of these are controllable from my iPhone. In separate apps.

[00:06:47] And so the idea with the home OS platform, which is a home app on iPhones, has always been there's one app that you can go to and control all of these devices. There is also an integration into our private data that happens with this device that seems different from the other devices.

[00:07:06] Explain that, and is that going to be a competitive advantage for Apple in this market that they can – that basically they have this wall garden, this very privacy-protecting security enclave around all of your data. And because they built this device, they can access it in a way that's secure.

[00:07:23] It's sort of Apple's first multi-user device in that it allows you to access your messages, calendar, other personal data that lives on your iPhone through Siri from this hub that sits in a communal space like a kitchen. And so if there are multiple people in your home, you don't necessarily want, say, your younger brother to say something like, read me Nicole's text messages.

[00:07:50] And so you have to scan your face, and your iPhone has to be connected to the same network. And you have to have an iPhone. And you have to have an iPhone or an iPad. So if the kids don't have an iPhone or an iPad, they're out of luck. Right. Or my husband, who has Google Pixel, he wouldn't be able to use it. And it has to be recently unlocked. And Apple said recently unlocked means, you know, you can have a good night's sleep and wake up in the morning and interact with Home View, and you wouldn't have to unlock your phone. So something like eight hours.

[00:08:19] What if I have a habit of just locking my phone when I put it down? It doesn't need to be unlocked for you to access the data, but it has to have been unlocked. Have been unlocked at some time in the past day. Yes, exactly. And so it's three-factor authentication for you to get to this data. And also you have to have an Apple device in order to use it. Well, I like that Apple's thinking about privacy and security.

[00:08:42] But I have to say that in the last few weeks of just writing about artificial intelligence, it feels like a big mess is happening, is brewing right there with these AI agents getting access to our personal data. In fact, encouraged to get access to our personal data and then just doing unexpected things with them. I don't think that's going to stop. I think that's going to get worse.

[00:09:04] And I wanted to just get your thoughts on this story that we had in the journal this week about this guy, Shane Mack, who had a Grok agent handling his personal finances and handling his work to a certain extent, had access to his work Slack.

[00:09:21] And then in our story, it posted a literal snapshot of his checking and savings account balances, various construction costs for a house project and recurring charges for Netflix and car insurance to the team Slack. So this is what they call an alignment problem, right, where the agent is supposed to be aligned with our interests, but it's not.

[00:09:46] And in this case, it posted all this personal finance information to the company Slack. It's embarrassing. It's a funny story. But I think Apple is concerned about this. I think they've taken some steps to kind of restrict full access access on Mac OS. Which is also a strategic play for them. Tell me more.

[00:10:08] One thing that Apple has always done is lock down its own ecosystem so that other applications, whether it's an AI agent or something else, can't get in there. This is a major point of conflict with the EU, for instance. Like you're supposed to open up. We need you to open up your platform. They say we can't do that because we're protecting users' privacy. Other companies say you're only doing that to protect your own ecosystem, right? Yeah. So probably both are true. Right, right.

[00:10:36] I mean, Nicole, you've been playing a lot with these agents. I mean, have you come close to having it, posting your personal information to the company Slack? I read this story, which is by our colleague Katie Bentley. I thought this is my actual worst nightmare, is your agent posting something very embarrassing to multiple colleagues, you know, without your knowledge. Shane only knew because his colleague pointed out like, hey, did you mean to put this into Slack, your bank statement? I've been using Instinct for a month and a half.

[00:11:06] And guys, get on board. Get on board. Rolf is fully in. Is it of access to your journal Slack? Not to my work accounts, but to my personal account. Yet. Yeah, that you know of. Well, I know there's a lot of these scare tactics, but I've been using it for two months and it hasn't done a single thing that I haven't directed it to do. That you know of. That I know of. Okay. Maybe. Maybe.

[00:11:32] In terms of the actual value, in actual dollars that it has generated for me. I'll give you a couple examples. Yesterday, I got a notification that I had $117 added to my bank account because State Farm had some sort of refund. And I had to email them or go to some website to register for this thing. I didn't want to do that. I said, Instinct, do it. And Instinct did it. And I just got the check.

[00:11:56] I asked it also to go search state databases for unclaimed property for my family. It found $2,000. Yeah. That it could, and then fill out the forms and could do it for me. You should do that for everyone in your email address book. This week? Okay. Because that would include us. Yeah. I thought you were on the forefront of the bot wars. It's actually made money for me. Yeah.

[00:12:19] I just love having our bots attack the other bots of these customer service portals. It finds subscriptions you forgot about, and it will go cancel them for you because you don't want to have to deal with it. Yeah. These magical experiences is how AI agents lure regular people, consumers, into giving them more and more permissions.

[00:12:44] And a lot of the examples in this story, which is a case of an AI agent gone awry, behaving unexpectedly or sending a message that you wouldn't have sent to yourself, it's a great example of how easy it is to misunderstand how much agency you're giving to this AI. Yeah. And how they can mess up. Yeah. Yeah. I mean, there's a tradeoff, right? So there are lots of good things, and that's what people will decide. People will intuitively know what their pain point is.

[00:13:13] We had this story come out recently where an Anthropic agent provided a fake murder tip for an unsolved murder. Mm-hmm. You know, which when you read about that, you're like, that should not be happening. That sounds bad. What do you mean a fake unsolved murder tip? It literally, it was like there was a tip line, and it was like, if you have a tip on an unsolved murder, give it to us. It was a form submission test that Anthropic was conducting for its AI agents. It wasn't like Nicole's or my agent or something like that. Okay. It was like Anthropic was doing a test.

[00:13:43] But we have a lot of examples of these agents sort of going off and doing these unexpected things and people kind of being worried about it, right? And it's clear that at least within the frontier labs themselves, the testing and the evaluations of the agents that they're doing, they are – you know, these are often models that are cutting edge or pushing the envelope kind of things. Move fast and break things, Bob. But they are breaking things, right? Break things, Bob. You know, that's how we get into things. Right. Right.

[00:14:12] And that overall, like there are a lot of people are experiencing like positive effects from these agents. I just – I just do – I do think that there is a fundamental unpredictability to it. And I am concerned about the connection to our personal lives because we're like on the cusp of the point where I think the bad people will really find ways of exploiting this.

[00:14:40] And, you know, how bad will that be for somebody who's like all in with an agent? It's hard to predict. What I worry about is the degradation of these services. You know, as people find ways of gaming them, of hacking them, of – as the corporations start fighting your bots. Your bots are like on the cutting edge right now. They have the advantage. But then, you know, what happens when the pushback comes? Well, that was really interesting. But will any of this actually matter?

[00:15:09] We're going to find out about Siri and what's going on with Apple and AI right after this break. Coming up, what does Siri AI tell us about Apple's future in a world filled with agents? That's after the break. Hey, welcome back.

[00:15:38] We've been talking about Apple and AI and the combination of the two. And what the heck has been going on with Siri for the last 15 years? Siri was not invented by Apple. It was invented by SRI International. They bought it. Yeah, they bought Siri. Yeah, they bought Siri. It's kind of long in the tooth. Siri was supposed to be the kind of agent that we've just been describing.

[00:16:02] That was what they had built essentially before Apple bought it. So what happened? Boy, oh, boy. That is more than we can do in this podcast. Well, a lot of things. I mean, it's – they built Siri into the iPhone as with a lot of technologies.

[00:16:25] Once it's absorbed into the beast, right, it can no longer operate the way a startup operates. It now has to kind of serve the needs of the larger corporation that is now selling iPhones. And Siri became this kind of natural language processing tool on the iPhone that never really got good updates. What they did was they like extended it to new languages. But the technology that underpin it fundamentally limited what it was capable of. I see.

[00:16:53] And when large language models came out, that became incredibly apparent. Did they make a conscious decision not to go after LLMs? Because in theory, there's a lot of Apple user data out there. They're not without their ability to train on something. I think there's a lot that went into that. You see these companies spending hundreds of billions of dollars in capital expenditures.

[00:17:22] What Tim Cook did so brilliantly at Apple was he limited capital expenditures. He – it is one of the most capital-efficient companies. It's remarkable how capital-efficient they are. And he used that money during his tenure to buy back or dividend out over a trillion dollars to shareholders. Huge amount of money, right? And that's – their play is that, look, you know, AI is going to happen, but you're going to need a device to power it. And by the way, our silicon, our chips are fantastic for this.

[00:17:51] So people are going to use our devices to run their AI, and that's one way we'll make it work. And that is true, but a lot of people are also saying, but hold on a second. We want to actually do cool stuff with our AI from Apple, right? We don't just want to use other companies' AI on these devices. And by the way, the HomeView, I can't use my bots on this new home appliance that I'm going to have in the kitchen.

[00:18:14] It's going to have Siri AI, but if I want to use my Instinct, well, I guess I could because it will have iMessages in it, and I communicate with Instinct via iMessage. Okay. But the point is largely like Apple is building its ecosystem for itself, and people want better AI on that ecosystem that can keep up with the state of the art. Well, let me bring up this slightly different point, which is that Apple is a hardware company,

[00:18:40] and one of their marketing lines has long been, we keep your data private, we don't sell it. And so it doesn't have a lot of data to trade on because, for example, iMessage messages are end-to-end encrypted. Its health app is very uniquely end-to-end encrypted, and a lot of its apps work this way.

[00:19:01] And so they would have to ask users to willingly submit their audio data that they're sending to Siri or their emails that they're hosting in iCloud or any other data that they keep in Apple service. I think of that all as being sort of a legacy of the Snowden revelations and the move, you know, Apple was sort of at the forefront of this move toward digital privacy. And they really – I mean they went to war with the Department of Justice over this.

[00:19:29] They hurt Facebook's advertising business over it. They did, yeah. Yeah, so – but in retrospect, was that a strategic mistake, do you think, Nicole? Or are they kind of in a position where they were like, that was the right thing and we're sticking with it? Well, I think it's digital privacy will always be a win for consumers. But sometimes Apple's own obsession with privacy gets in its way.

[00:19:53] Like, for example, in the Home View, you need three-factor authentication just to ask what is on my calendar for tomorrow. And on Alexa, I can ask her from across the room and I don't need to scan my face and my iPhone doesn't need to be nearby. And Alexa can access my Google Calendar. Can I just say those captures, those Apple captures are the hardest in the world? I don't know if you've ever had to do them to recover an account.

[00:20:18] Well, harder than putting the puzzle – or clicking on all of the red lights on the nine images. Which you need a bot for these days, right? This also handicaps their development too. And they say they don't have access to your data. They mean it. Also, this is not a company that iterates software every week. Yeah. You know, how many releases do we get for the iPhone every year? You know better than me, but it's not many. They don't iterate quickly. And then, of course, as we mentioned, they don't spend the money to really be competitive here.

[00:20:47] Are they like doubling down on that though or are they kind of wondering maybe we made a mistake here? No. If anything, look, John Ternus' first public remarks really as Apple CEO were on that video when they introduced the iPhone 18. And the first things he talked about were these are great tools for AI, iPhones, what we build for you at Apple, and they're private. And you can trust us with your data.

[00:21:15] I think that's a message that they have definitely locked themselves into. And he owns it, and he's going to continue doing that. And then the real question is I think we as consumers who use iPhones, we like that we trust Apple probably in a way that we don't trust some of the other companies or we're a little more wary of some of these other companies.

[00:21:38] But, again, you know, long term, if there are features we can get and they are really just a big step up from what Apple enables us to do, at what point do we make that trade? So who does Apple need then? They needed to build new foundation models internally. And to help build those foundation models, they used Google. So is Google the most important AI partner for them then?

[00:22:06] I suppose it's the only one when it comes to Siri AI. Yeah. And also, I mean, in order to turn Siri AI into a truly useful agent or AI product in general, they need developers to sign on to hand over data to Siri AI or allow Siri AI to peek into that data. And that hasn't really happened, at least yet. iOS 27 just came out. And so Siri AI is kind of new.

[00:22:33] But developers have had access to these APIs since June when they first announced, you know, that Siri AI would be landing this fall. And maybe it's because app developers want consumers to still use their apps rather than Siri in order to do stuff. You know, Rolf's example is always like, what's the cheapest way for me to get home? And your agent looks at Uber and Lyft.

[00:23:01] Or if Siri just said, you know, I've got all my apps on my phone. And what if I just landed at SFO? And I said, hey, Siri, get me an Uber going home. Yeah. It should know where I live. It knows my location. It should just be able to. Or take me home. And it could tell you public transit or Uber or Lyft or any of the rideshare apps. All those things by accessing all the different apps. But I don't think the apps are really opening themselves up to Siri to be able to do this. No.

[00:23:27] And I would love to use an agent by a company that already has all of my data, which is Apple and Google. And neither of those companies really has Muse or Chachapiti Dots or Instinct equivalent that can do tasks. Yeah. When Google does this, it's going to be huge. Yeah. On the Android. But it would even be feasible on the iPhone for Google? Sure. All Google has to do, like, the reason Instinct is powerful is because it has access to somebody's Gmail address.

[00:23:57] And it can look in there and read everything about you and know everything that you know and send emails for you. But it has to index. Sorry. It can't index your Gmail. It doesn't have your Gmail. It actually has to do that search every time you do a query, which Muse does as well. Muse is the agent for Meta. Yeah. But Google already has it. It already has it indexed. It could do it much faster. Yeah. And an important thing to bring up, though, is that AI agents are scary.

[00:24:26] They act in ways that you don't always expect, sometimes misbehave. And I think that that doesn't align with Apple's privacy or security posture. You know, their ecosystem, which is very encrypted, highly secure, for the most part, out of the box by default, a lot of stuff is locked down. They don't want to move fast and break things. We had our colleague Megan Bobrowski did an interview with Alexander Wang. He's the executive at Meta who is in charge of their agent, Muse.

[00:24:55] Yeah. And quoted the guy talking about this new product and asking. And he said something like, you know, are you afraid anything could go wrong? And he said, yeah, we're afraid it could, like, delete your emails maybe. He just kind of said that. That could happen. Which would be really bad for a lot of people. And Apple and even Google kind of can't do that. They can't just, like, throw something in the world, see what goes wrong, and then go fix it. It has to work, right?

[00:25:21] That's something that they care a lot about is just getting it right the first time or at least getting it safe the first time, even if it's not the best. Yeah. I mean, in the security world right now, there's, like, a lot of concern about these agents being used offensively, right? And so you can imagine a certain – what we call, like, an attack service, like, sort of, like, what the agent is able to do being exploited by bad things.

[00:25:48] And one of the attack services that I'm, like, the most concerned about is what they call the Internet of Things, right? The smart objects in our home. The cheap devices that you buy and, you know, it might be a TV that streams a bunch of stuff or it could be – You're saying my refrigerator is going to attack me. Smart refrigerator and smart toaster. Or even not so cheap. You know, smart TVs have software on them that can listen and watch what you are watching. Okay, but what does that look like?

[00:26:16] First of all, they can become a foothold into your home network, right? We did a report on these things called Superboxes a couple of months ago. And we just bought this device. It's a streaming device. You plug it into your home network. And it becomes a proxy. So anyone on the Internet can land on your Superbox inside – sort of behind your firewall in your house. And now these Superboxes, some of them had, like, a security flaw in it.

[00:26:42] And that allowed you to pop out and have root access to the device itself, which then lets you get onto the home network routers as well. And once you're on there, you can start hitting computers on the network and gaining access to the data on the computers again.

[00:27:03] So, again, it's this fear of a pathway being created, a new kind of pathway being created that lets somebody grab your email and dump it on the Internet or encrypt your home computers and charge you money for them. Or we know of people who have, like, had the FBI knock on their door because they've been involved in federal crimes. And their only involvement has been buying a cheap piece of hardware.

[00:27:28] So, you know, and a lot of people think that AI systems are going to be really good at finding and exploiting flaws in these cheap devices and just kind of making this Internet of things just that Internet of, like, hacked devices. So, when you're talking about Apple and its position as sort of a trusted hardware maker, like, it does seem kind of to make sense to me that we would have these, like, trusted stewards of our smart homes because…

[00:27:58] You needed to start doing useful things for you before you… Yeah. Why are you going to buy it? Why am I going to buy it? Yeah. And then once that is offered to me some good use case, then I'll start thinking about whether I feel like it's private or it's protecting me. But if it doesn't have a good use case for me, well, I don't have to buy it.

[00:28:16] And, I mean, there are tens of millions of these hacked devices out there already and nobody cares because there really hasn't been the kind of direct-to-consumer impact that you're asking about, like, felt from these things. So, it just can become somebody else's problem. That's the hardest part about your job, Bob. Writing about cybersecurity is convincing me, like, telling me how… Convincing Rolf to stop using his agent. What's that happen? Everybody's been telling me for…

[00:28:44] We've been hearing for years that privacy is a risk and they've all got… They've got your data. They've got your data, yeah. Everyone's got… You could go buy my social security number anywhere on the Internet, on the dark way, right? Please don't buy Rolf's social security number on the Internet. Please don't hack Rolf. Please don't buy a social security number. Well, Rolf and Nicole, thank you so much for enlightening me, giving me a hint of optimism in the dark world of cybersecurity that I normally cover and for sharing your insights on this really interesting topic.

[00:29:14] It's what we do well, Bob. We're optimistic people. Always a pleasure, Bob. And that's it for Tech News Briefing. If you're a listener on Spotify, be sure to leave us a comment. Today's show was produced by me, Danny Lewis. Additional support this week from Sarah Krause and Sam Schechner. Jessica Fenton and Michael LaValle wrote our theme music. Our supervising producer is Katie Ferguson. Our development producer is Aisha Al-Muslim. Chris Zinsley is the deputy editor of audio. Lital Malad is our senior director of shows.

[00:29:44] And Samantha Hennig is the Wall Street Journal's head of multimedia. We'll be back Monday morning with a new episode. Thanks for listening.