The first big fight for life after the smartphone

Apple’s new lawsuit against OpenAI is not just about whether a few former employees walked out with trade secrets. It is one of the first visible battles over who controls computing in a world where AI agents, not apps, sit between people and the digital systems they use every day.

For more than a decade, the smartphone has been the organising centre of personal computing. We live inside it. We tap icons, switch between apps, juggle notifications and authenticate transactions. But every major technology era has ended the same way: not with a formal announcement, but with leading companies quietly stepping beyond the boundaries that once defined them. Mainframes did not vanish when the PC arrived; they simply stopped defining what “computing” meant. Newspapers, retailers and travel agents did not close the day the web launched; they lost their central position over time. The smartphone, too, is unlikely to disappear overnight. What is at risk is its role as the place where digital interactions begin.

Seen in that light, Apple vs OpenAI is not just a legal case. It is a test of what comes after the app-and-screen era.

Apple’s complaint, filed in the Northern District of California on July 10, 2026, accuses two former employees and OpenAI-related entities of misappropriating confidential information. According to public reporting, Apple claims these individuals accessed internal systems without permission, downloaded sensitive materials, discussed unreleased projects in hiring conversations and were encouraged to bring physical parts or design artefacts to interviews. These are allegations, not proven facts, and they will be tested in court. But the breadth of what Apple describes is revealing. The company does not frame its secrets as a single component or one new product. It talks about entire systems: hardware architecture, manufacturing processes, specialised equipment, power and thermal management, supplier relationships and the finely tuned coordination required to ship new devices at global scale.

That scope tells us what Apple thinks is truly at stake: not just a design, but an operating model for building physical computing platforms that can support the next era.

The deeper question: who owns the agent?

The legal arguments will revolve around contracts, access logs and trade-secret law. Beneath that, the strategic question is simpler and far more consequential:

Who will own the AI agent that sits between human intention and digital action?

In the smartphone era, your intent flows like this:

• You decide what you want to do

• You pick up your phone

• You choose an app

• You follow that app’s steps to get to an outcome

A simple request such as, “Move my trip to Tuesday, keep the customer meetings, find a comparable hotel and tell everyone affected,” is scattered across multiple apps and services. You open the airline app, adjust the booking, find a hotel site, update your calendar, message participants and approve payments. You are the orchestration layer. In the agentic era, that structure flips. You express an outcome; the agent orchestrates the rest. It interprets your goal, checks your constraints, chooses tools, talks to systems and executes the workflow inside your permissions.

The phone is still present, but no longer necessarily in charge. It becomes one of several channels through which an agent sees the world and takes action. Owning that agent—the default layer through which people express intentions and let things happen—could turn out to be more powerful than owning any single app or even any single device.

On paper, Apple and OpenAI looked like natural partners. Apple controlled the hardware, operating system, identity, security model and customer relationship. OpenAI brought frontier models and a rapidly evolving AI stack. Apple could keep the user experience coherent while quietly invoking OpenAI’s intelligence when needed. That arrangement worked in the first wave of generative AI, when models mostly answered questions and stayed behind familiar interfaces. But it was never likely to remain stable once AI began to act on a user’s behalf.

A model provider does not want to live forever behind someone else’s operating system. It eventually wants direct access to context, permissions to act, and a path to transactions and ongoing relationships. A device company, in turn, cannot be comfortable if a third party’s intelligence becomes more important than the device itself. The tension is tolerable when AI is a feature. It becomes existential when AI is the new interface.

OpenAI’s hardware push changes the game

That is why OpenAI’s move into hardware matters so much. The company is working with former Apple design chief Jony Ive and other experienced hardware leaders on new consumer devices, including a smart speaker and related products designed around native AI capabilities rather than traditional apps.These projects are not side experiments. They are an effort to control the post-smartphone interface. 

The smartphone is app-centric. Users select an application, learn its interface and click through a series of steps. A truly AI-native device could be intent-centric. Users state what should happen; the system chooses tools, services and data sources, and quietly completes the work.

In that world:

• The model provides reasoning.

• The device provides presence.

• The operating system provides authority.

• Identity systems provide permission.

• Applications and services provide execution.

• Payments provide economic agency.

That stack makes the AI agent tangible. It lives in a device, has a continuous sense of context and can act within well-defined guardrails. The company that ties these layers together does not just ship a better assistant. It becomes a gatekeeper for how people interact with the digital and physical world around them. Apple and OpenAI are racing toward the same centre from opposite directions. Apple starts with devices and is pushing “up” into intelligence. OpenAI starts with intelligence and is pushing “down” into devices and operating environments. Their paths were always going to collide.

The post-smartphone world: survival vs centrality

The important question is not whether OpenAI can out-design Apple and produce a “better” phone. The question is whether the smartphone itself remains the organising centre of everyday computing. History suggests that mature technologies rarely vanish. Mainframes, radio and television all survived later waves. They just stopped defining their categories. A more useful metaphor is the transition from horses to automobiles. Early cars were expensive, unreliable and poorly supported; roads, fuel stations and repair networks took years to develop. For a long time, horses and cars coexisted. What changed was not the existence of horses, but their role in the transportation system.

The same pattern could play out with smartphones.

• The phone remains an excellent camera, secure identity device, communications hub and high-quality display.

• It may live in our pockets for many years.

• Yet it may gradually stop being the place where digital journeys begin.

Instead of asking, “Which app should I open?” people may increasingly state what they want to happen. The agent will decide which apps, merchants, services, devices and payment systems to use. The phone will still participate, but more as secure infrastructure beneath the agent than as the main stage. In strategic terms, the phone survives physically but recedes from the foreground of the digital economy.

For Apple, that is far more threatening than another smartphone competitor. The iPhone’s power does not come only from hardware sales. It comes from being the front door: controlling identity, app discovery, payments, notifications, sensors and access to services. If an external agent becomes the primary interface through which users express needs and delegate action, Apple risks being pushed one layer down the stack.

The device would still be in your hand. The control point would move above it.

Why Apple’s execution machine matters

Apple’s complaint underscores something the industry often overlooks: the difference between invention and execution. A small team can build a compelling prototype of a new device or agent. Turning that into millions of reliable, secure products is a different discipline entirely. It demands deep expertise in batteries, thermals, materials, acoustics, tolerances, logistics, failure rates and supplier ecosystems.

A chef can invent an extraordinary dish in a single kitchen. Building a global chain that reproduces that experience every day requires a different operating system: sourcing, training, equipment, quality control, distribution and culture. Apple’s advantage has never been just in the “recipe” of a single product. It lies in the “kitchen” it has built: the supply chains, factories, processes and disciplines that let it deliver integrated devices and services at planetary scale.

For any AI company moving into hardware, the journey from intelligence to physical execution is long. A powerful model is not a finished product. A captivating demo is not a manufacturing system. A conversational interface is not a device that can run 24/7 in the messy real world. That is why Apple’s institutional knowledge is so sensitive—and why, if Apple’s allegations are proven, it would see any leakage as an existential threat rather than a narrow IP problem.

The physical agentic stack

In earlier discussions of Agentic AI, the industry has drawn a line between conversational systems and operational systems. Generative AI makes intelligence accessible through language. Agentic AI makes that intelligence capable of pursuing goals, invoking tools, handling multi-step tasks and completing work within defined boundaries.

What Apple vs OpenAI highlights is that this “agentic” concept is now extending into physical computing.

For an agent to be truly useful in daily life, it needs:

• Identity, so it can prove who it is acting for

• Permissions and policies, so it knows what it is allowed to do

• Memory and context, so it can act consistently over time

• Tools and integrations, so it can actually change the world, not just describe it

• Hardware sensors, processors, displays and connectivity, so it can see, hear, speak and respond in real environments

In other words, the agentic stack is not only digital. It is physical.

The players who can integrate models, devices, identity, payments and services into a coherent, trustworthy system will not just ship gadgets. They will shape the norms and economics of the post-smartphone era.

Talent, boundaries and trust

The case also surfaces a long-standing tension in Silicon Valley: where professional experience ends and proprietary knowledge begins. The region has always thrived on mobility. Engineers move between companies, founders leave incumbents to start new ventures, and knowledge spreads through networks. But the AI race is tightening the screws. When an AI company wants to build hardware, it naturally targets people who have already done batteries, cameras, supply chains and factories. Likewise, when a device company doubles down on AI, it wants model researchers, inference engineers and agent architects.

The line is crossed when hiring is not just about skills, but about extracting internal methods, supplier details, confidential roadmaps or proprietary processes. Apple’s complaint describes recruitment conversations in which candidates allegedly discussed sensitive project details and were asked to bring physical components or design artefacts.

Regardless of how this specific case is decided, it signals the need for stronger governance:

• Interviewers must be trained to stop candidates from disclosing confidential information.

• Technical assessments should emphasise problem-solving over showcasing proprietary work.

• New hires should receive explicit guidance on what knowledge they can and cannot use.

• In sensitive areas, companies may need clean-room documentation to prove independent development.

As agents gain more autonomy, the provenance of models, data and even hardware design will become a central trust issue, not a back-office compliance task.

Beyond intelligence: trust as architecture

Agentic systems carry a different risk profile from chatbots. A chatbot can be wrong; an agent can take the wrong action.

It can move money, reschedule a surgery, cancel a flight, notify a client or expose sensitive data. Once intelligence is allowed to act, trust stops being a marketing message and becomes part of the system architecture.

That architecture includes:

• Identity and authentication

• Fine-grained permissions and policy controls

• Audit logs and explainability

• Human escalation paths and revocation mechanisms

• Clear data boundaries and minimisation practices

Apple has built much of its brand on secure devices, privacy and tight integration. OpenAI’s strengths lie in model capability and rapid product iteration with broad adoption. Both approaches now meet at the same point: this is about winning the trust this entity to be your default agent

A larger platform war opening up

Zooming out, the Apple–OpenAI conflict is a preview of broader shifts.

• Partnerships will become more unstable, as companies cooperate at one layer (say, devices and models) while competing at another (interfaces and agents).

• Economic power will move from attention to intention: from capturing clicks in apps to interpreting goals and executing on behalf of users.

• The most defensible positions will sit at the intersections: between models and devices, devices and identity, identity and payments, agents and services.

Consumer markets are only one front. Inside enterprises, the same race is underway among cloud providers, SaaS platforms, model companies and services firms, each trying to become the orchestration layer through which work flows. Whoever owns business logic and the experience layer for agents will wield outsized influence over the “agentic enterprise.”

In this sense, Apple’s lawsuit is not just a reaction to a specific set of employees or a single partnership. It is one of the first shots in a larger platform war, where the real prize is control of the interface between human intention and machine action in a post-smartphone era. The smartphone will likely remain in our pockets. But the battle now is over what sits on top of it—and who gets to choreograph the agents that increasingly run our digital lives.