This week, OpenAI announced that apps can now run directly inside ChatGPT, letting users book travel, create playlists, and edit designs without switching between different apps. Some immediately declared the app platform of the future — predicting a ChatGPT-powered world where Apple’s App Store becomes obsolete.

But while OpenAI’s app platform presents an emerging threat, Apple’s vision for an improved Siri — though still seriously delayed — could still play out in its favor.

After all, Apple already controls the hardware, the operating system, and has roughly 1.5 billion iPhone users globally, compared to ChatGPT’s 800 million weekly active users. If Apple’s bet pays off, it could position the iPhone maker in a way that would not only maintain its app industry dominance but also modernize how we use apps in the AI era.

Apple’s plan is to kill the app icon without killing the app itself. Its vision for AI-powered computing — introduced at its developer conference last year — would see iPhone users interact with an overhauled version of Siri and a revamped system that changes the way you use apps on your phone. (Imagine less tapping and more talking.)

Apps are passé, long live apps?

It’s an idea whose time has come.

Organizing little tappable icons on your iPhone’s Home Screen to make online information more accessible is a dated metaphor for computing. Meant to resemble a scaled-down version of a computer’s desktop, apps are becoming a less common way for users to interact with many of their preferred online services.

These days, consumers are just as likely to ask an AI assistant for a recommendation or insight as they are to do a Google search or launch a dedicated, single-purpose app, like Yelp. They’ll talk out loud to their smart speakers or Bluetooth-connected AirPods to play their favorite tunes; they’ll ask a chatbot for business information or a summary of reviews for a new movie or show.

The AI, a large language model trained on web-scraped data and more, determines what the user wants to know and spits out a response.

This is arguably easier than scouring through Google’s search results for the right link with the answer. (That’s something Google itself realized over a decade ago, when it started putting answers to user queries right on the search results page.)

AI is also often easier than finding the right app on your now overcrowded iPhone, launching it, and then interacting with its user interface — which varies from app to app — to perform your task or get an answer to your question.

However, ChatGPT’s app system, while seemingly improving on this model, remains locked inside the ChatGPT user experience. It requires consumers to engage in a chatbot-style interface to use their apps, which could require user education. To call up an app, you have to name it as the first word of your prompt or otherwise mention the app by name to get a button that prompts you to “use the app for the answer.” Then, you have to type in an accurate query. (If you mess this up, early tests by Bloomberg indicate you could get stuck on a loading screen with no results!)

We have to wonder: is this the future of apps, or just the future while there’s no other competition? When another solution becomes available — one that’s built into your iPhone, no less — will consumers keep using ChatGPT, or are they still willing to give Siri another try? We don’t know, but we wouldn’t count out Apple yet, even though Siri has quite a bad reputation to salvage at this point.

Siri may be an embarrassment as it stands today, but Apple’s overall ecosystem has advantages. For starters, consumers already have the apps they want to use on their phone or know how to find them on the App Store, if not. They’ve used many of these apps for years. Muscle memory goes a long way!

Meanwhile, there are a few roadblocks to getting started with ChatGPT’s app platform.

You have to install the app in question, of course; then you have to connect the app to ChatGPT by jumping through a warning-filled permission screen. This process requires you to authenticate with the app using your existing username and password, and to enter the two-factor authentication code, if applicable.

After this one-time setup, things should be easier. For instance, after you generate a Spotify playlist with AI, it can be launched in the Spotify app with a tap.

However, this experience won’t differ much from Apple’s plans if Apple is able to make things work as promised. Apple says you’ll be able to talk or text Siri to control your apps.

There are other disadvantages to the OpenAI app model. You can only interact with one app at a time, instead of being able to switch back and forth between apps — something that could be useful when comparing prices or trying to decide between a hotel room and an Airbnb.

Using apps within ChatGPT also strips away the branding, design, and identity that consumers associate with their favorite apps. (For those who hate how cluttered Spotify’s app has become, perhaps that’s a good thing. Others, however, will disagree.) And, in some cases, using the mobile app version to accomplish your goals may still be easier than using the ChatGPT app version because of the flexibility the former offers.

Finally, compelling users to switch app platforms could be difficult when there isn’t an obvious advantage to using apps within ChatGPT — except for the fact that it’s neat that you can.

Can Apple save Siri’s reputation with AI features?

In its WWDC 2024 demonstration — which Apple swears was not “demoware” — the company showed how the apps would function under this new system and how they could use other AI features like proofreading.

Most importantly, Apple told developers that they’ll be able to take advantage of some of its AI capabilities without having to do additional work — like a note-taking app using proofreading or rewriting tools. Plus, developers who have already integrated SiriKit into their apps will be able to do more in terms of having users take action in their apps. (SiriKit, a toolkit for making apps interoperable with Siri and Apple’s Shortcuts, is something developers have been using since iOS 10.)

These developers will see immediate enhancements when the new Siri rolls out.

Apple said it will focus on categories like Notes, Media, Messaging, Payments, Restaurant Reservations, VoIP Calling, and Workouts, to start.

Apps in these categories will be able to let their users take actions via Siri. In practice, that means Siri will be able to invoke any item from an app’s menus. For example, you could ask Siri to see your presenter notes in a slide deck, and your productivity app would respond accordingly.

The apps would also be able to access any text displayed on the page using Apple’s standard text systems. That could make the app interactions feel more natural, without the user having to give specifically worded prompts or commands. For instance, if you had a reminder to wish your grandpa a happy birthday, you could say “FaceTime him” to take that action.

Apple’s existing Intents framework is also being updated to gain access to Apple Intelligence, covering even more apps in categories like Books, Browsers, Cameras, Document Readers, File Management, Journals, Mail, Photos, Presentations, Spreadsheets, Whiteboards, and Word Processors. Here, Apple is creating new “Intents” that are pre-defined, trained, and tested, and making them available to developers.

That means you could tell the photo-editing app Darkroom to apply a cinematic filter to an image via Siri. Plus, Siri will be able to suggest an app’s actions, helping iPhone users discover what their apps can do and take those actions.

Developers have been adopting the App Intents framework, introduced in iOS 16, because it offers other functionality to integrate their app’s actions and content with other platform features, including Spotlight, Siri, the iPhone’s Action button, widgets, controls, and visual search features — not just Apple Intelligence.

Also, unlike ChatGPT, Apple runs its own operating system on its own hardware and offers the App Store as a discovery mechanism, the app infrastructure, and developer tools, APIs, and frameworks — not just the AI-powered interface that will help you use your apps.

Though Apple may have to borrow some AI tech from others to do that last bit, it has the data to personalize your app recommendations, and, for the privacy-minded, the controls that let you limit how much information apps themselves can collect. (Where’s the “Do Not Track” option for ChatGPT’s app system, we wonder?)

OpenAI’s system doesn’t work out of the box with all your apps at launch. It requires developer adoption and relies on the Model Context Protocol (MCP), a newer technology for connecting AI assistants to other systems. That’s why ChatGPT currently works with only a handful of apps, like Booking.com, Expedia, Spotify, Figma, Coursera, Zillow, and Canva. MCP adoption is growing, but the delay in its becoming broadly adopted could give Apple the extra time it needs to catch up.

What’s more, word is that Apple’s AI system is nearly ready. The company is reportedly already internally testing this, allowing users to take actions in apps by using Siri voice commands. Bloomberg reported that this smarter version of Siri works out of the box works with many apps, including those from major players like Uber, AllTrails, Threads, Temu, Amazon, YouTube, Facebook, and WhatsApp. And it’s still on track to ship next year, Apple confirmed to TechCrunch.

Apple has an iPhone, OpenAI has Jony Ive

The iPhone’s status as an app platform will also be difficult to disrupt, even from a company as large and powerful as OpenAI.

The ChatGPT maker understands this, too, which is why OpenAI is exploring its own device with Apple’s former head of design, Jony Ive. It wants its AI to become more of a part of consumers’ everyday lives and habits, which could require a hardware device.

But, so far, the company has struggled to think up a better computing paradigm than the smartphone, reports indicate. At the same time, the general public has demonstrated an aversion to always-on AI devices, which bump up against existing social norms and threaten privacy.

The AI backlash has covered AI device maker Friend’s NYC subway posters, led Taylor Swift fans to attack their idol for dabbling in AI, and threatened the reputation of popular consumer brands and enterprise businesses alike. That leaves the future success of an OpenAI device in question.

For now, that means OpenAI’s app model is one that essentially boils down to using its app to control other apps.

If Apple gets its Siri upgrade right, that intermediary may not be necessary.

Chris Lehane is one of the best in the business at making bad news disappear. Al Gore’s press secretary during the Clinton years, Airbnb’s chief crisis manager through every regulatory nightmare from here to Brussels – Lehane knows how to spin. Now he’s two years into what might be his most impossible gig yet: as OpenAI’s VP of global policy, his job is to convince the world that OpenAI genuinely gives a damn about democratizing artificial intelligence while the company increasingly behaves like, well, every other tech giant that’s ever claimed to be different.

I had 20 minutes with him on stage at the Elevate conference in Toronto earlier this week – 20 minutes to get past the talking points and into the real contradictions eating away at OpenAI’s carefully constructed image. It wasn’t easy or entirely successful. Lehane is genuinely good at his job. He’s likable. He sounds reasonable. He admits uncertainty. He even talks about waking up at 3 a.m. worried about whether any of this will actually benefit humanity.

But good intentions don’t mean much when your company is subpoenaing critics, draining economically depressed towns of water and electricity, and bringing dead celebrities back to life to assert your market dominance.

The company’s Sora problem is really at the root of everything else. The video generation tool launched last week with copyrighted material seemingly baked right into it. It was a bold move for a company already getting sued by the New York Times, the Toronto Star, and half the publishing industry. From a business and marketing standpoint, it was also brilliant. The invite-only app soared to the top of the App Store as people created digital versions of themselves, OpenAI CEO Sam Altman; characters like Pikachu, Mario, and Cartman of “South Park”; and dead celebrities like Tupac Shakur.

Asked what drove OpenAI’s decision to launch this newest version of Sora with these characters, Lehane gave me the standard pitch: Sora is a “general purpose technology” like electricity or the printing press, democratizing creativity for people without talent or resources. Even he – a self-described creative zero – can make videos now, he said on stage.

What he danced around is that OpenAI initially “let” rights holders opt out of having their work used to train Sora, which is not how copyright use typically works. Then, after OpenAI noticed that people really liked using copyrighted images, it “evolved” toward an opt-in model. That’s not really iterating. That’s testing how much you can get away with. (And by the way, though the Motion Picture Association made some noise last week about legal threats, OpenAI appears to have gotten away with quite a lot.)

Naturally, the situation brings to mind the aggravation of publishers who accuse OpenAI of training on their work without sharing the financial spoils. When I pressed Lehane about publishers getting cut out of the economics, he invoked fair use, that American legal doctrine that’s supposed to balance creator rights against public access to knowledge. He called it the secret weapon of U.S. tech dominance.

Maybe. But I’d recently interviewed Al Gore – Lehane’s old boss – and realized anyone could simply ask ChatGPT about it instead of reading my piece on TechCrunch. “It’s ‘iterative’,” I said, “but it’s also a replacement.”

For the first time, Lehane dropped his spiel. “We’re all going to need to figure this out,” he said. “It’s really glib and easy to sit here on stage and say we need to figure out new economic revenue models. But I think we will.” (We’re making it up as we go, in short.)

Then there’s the infrastructure question nobody wants to answer honestly. OpenAI is already operating a data center campus in Abilene, Texas, and recently broke ground on a massive data center in Lordstown, Ohio, in partnership with Oracle and SoftBank. Lehane has likened accessibility to AI to the advent of electricity – saying those who accessed it last are still playing catch-up – yet OpenAI’s Stargate project is seemingly targeting some of those same economically challenged places as spots to set up facilities with their massive appetites for water and electricity.

Asked during our sit-down whether these communities will benefit or merely foot the bill, Lehane went to gigawatts and geopolitics. OpenAI needs about a gigawatt of energy per week, he noted. China brought on 450 gigawatts last year plus 33 nuclear facilities. If democracies want democratic AI, they have to compete. “The optimist in me says this will modernize our energy systems,” he’d said, painting a picture of re-industrialized America with transformed power grids.

It was inspiring. But it was not an answer about whether people in Lordstown and Abilene are going to watch their utility bills spike while OpenAI generates videos of John F. Kennedy and The Notorious B.I.G. (Video generation is the most energy-intensive AI out there.)

Which brought me to my most uncomfortable example. Zelda Williams spent the day before our interview begging strangers on Instagram to stop sending her AI-generated videos of her late father, Robin Williams. “You’re not making art,” she wrote. “You’re making disgusting, over-processed hotdogs out of the lives of human beings.”

When I asked about how the company reconciles this kind of intimate harm with its mission, Lehane answered by talking about processes, including responsible design, testing frameworks, and government partnerships. “There is no playbook for this stuff, right?”

Lehane showed vulnerability in some moments, saying that he wakes up at 3. a.m. every night, worried about democratization, geopolitics, and infrastructure. “There’s enormous responsibilities that come with this.”

Whether or not those moments were designed for the audience, I believe him. Indeed, I left Toronto thinking I’d watched a master class in political messaging – Lehane threading an impossible needle while dodging questions about company decisions that, for all I know, he doesn’t even agree with. Then Friday happened.

Nathan Calvin, a lawyer who works on AI policy at a nonprofit advocacy organization, Encode AI, revealed that at the same time I was talking with Lehane in Toronto, OpenAI had sent a sheriff’s deputy to his house in Washington, D.C., during dinner to serve him a subpoena. They wanted his private messages with California legislators, college students, and former OpenAI employees.

Calvin is accusing OpenAI of intimidation tactics around a new piece of AI regulation, California’s SB 53. He says the company weaponized its legal battle with Elon Musk as a pretext to target critics, implying Encode was secretly funded by Musk. In fact, Calvin says he fought OpenAI’s opposition to California’s SB 53, an AI safety bill, and that when he saw the company claim it “worked to improve the bill,” he “literally laughed out loud.” In a social media skein, he went on to call Lehane specifically the “master of the political dark arts.”

In Washington, that might be a compliment. At a company like OpenAI whose mission is “to build AI that benefits all of humanity,” it sounds like an indictment.

What matters much more is that even OpenAI’s own people are conflicted about what they’re becoming.

As my colleague Max reported last week, a number of current and former employees took to social media after Sora 2 was released, expressing their misgivings, including Boaz Barak, an OpenAI researcher and Harvard professor, who wrote about Sora 2 that it is “technically amazing but it’s premature to congratulate ourselves on avoiding the pitfalls of other social media apps and deepfakes.”

On Friday, Josh Achiam – OpenAI’s head of mission alignment – tweeted something even more remarkable about Calvin’s accusation. Prefacing his comments by saying they were “possibly a risk to my whole career,” Achiam went on to write of OpenAI: “We can’t be doing things that make us into a frightening power instead of a virtuous one. We have a duty to and a mission for all of humanity. The bar to pursue that duty is remarkably high.”

That’s . . .something. An OpenAI executive publicly questioning whether his company is becoming “a frightening power instead of a virtuous one,” isn’t on a par with a competitor taking shots or a reporter asking questions. This is someone who chose to work at OpenAI, who believes in its mission, and who is now acknowledging a crisis of conscience despite the professional risk.

It’s a crystallizing moment. You can be the best political operative in tech, a master at navigating impossible situations, and still end up working for a company whose actions increasingly conflict with its stated values – contradictions that may only intensify as OpenAI races toward artificial general intelligence.

It has me thinking that the real question isn’t whether Chris Lehane can sell OpenAI’s mission. It’s whether others – including, critically, the other people who work there – still believe it.