Something shifted this week — quietly, almost bureaucratically — and I think it matters more than the headlines around it suggested.
Apple filed a lawsuit against OpenAI alleging trade secret theft, citing the acquisition of a company called IO Products as the mechanism. On the same day, Apple confirmed that the new Siri — launching this autumn — will use Google Gemini as its underlying model, not ChatGPT. The two announcements arrived together, and the combination is worth sitting with for a moment. Because this is not just a corporate dispute or a product pivot. It is a signal that the foundational alliances of the current AI era are dissolving, and that what replaces them is going to be stranger and more contentious than the straightforward competition everyone assumed was coming.
Go back eighteen months. The deal between Apple and OpenAI felt, at the time, like a landmark. Apple — historically the most reluctant of the major tech platforms to depend on outside intelligence — was embedding ChatGPT directly into iOS. The framing from both sides was careful: partnership, integration, complementary strengths. What it actually was, in retrospect, was a temporary détente between two companies that were always going to end up as adversaries. Apple needed AI capabilities faster than it could build them. OpenAI needed Apple's distribution. Each was using the other. The cooperation was real, but it was also a stopgap, and both sides presumably knew it.
What I find more interesting than the lawsuit itself is the Siri-to-Gemini switch. Think about the geometry of that decision. Apple spent a year building consumer trust around the OpenAI integration — millions of people used ChatGPT through Siri, or at least were told they could. And now, without any particular fanfare, the whole stack is moving to Google. The model powering the most intimate AI assistant on the most widely used smartphone in the world just changed hands. And most of those users will not notice, because most of them don't think in terms of which model is running underneath. They just think of it as Siri.
That invisibility is itself the story. One of the things the AI industry has been quietly negotiating for the past three years is whether the model or the interface is the thing that earns user loyalty. The answer — which Apple's move makes concrete — is the interface. It almost always is. People don't switch phones because the underlying language model changed. They stay because the camera is good and their apps are there and changing is annoying. The model is infrastructure. The interface is the relationship. And if that is true, it changes what the competition is actually over.
It means that OpenAI's real vulnerability is not being outperformed on benchmarks by Claude or Grok. The benchmarks are important but they are not the moat. The moat is being embedded so deeply in the workflows and interfaces that people use every day that switching feels like friction. And Apple just demonstrated, rather elegantly, that it can swap one model for another with essentially no user-facing disruption. Which means the moat OpenAI thought it had in the iOS ecosystem was shallower than it appeared.
This is what I keep turning over. The AI companies spent 2023 and 2024 in a kind of cooperative frenzy — partnerships with Microsoft, Apple, Samsung, a dozen enterprise platforms. The logic was: distribute widely, get into everything, earn the switching costs before the competition properly begins. And it worked, in the sense that everyone is now everywhere. But it also created a world where the platforms — the Apples and Microsofts and Googles — have had two years to observe exactly how these models perform, exactly where their weaknesses are, and exactly how much leverage they hold as distribution partners. The observation window gave them power. And now some of them are starting to use it.
The lawsuit complicates this further. Trade secret cases between technology giants are common enough to be almost unremarkable as a category — but this one lands at a particular moment. OpenAI is in its IPO window. Active litigation from one of the most valuable companies in the world is a material risk disclosure that goes in the S-1. It creates uncertainty for investors at exactly the point when OpenAI most needs to project stability and inevitability. Whether that timing is intentional — whether Apple calculated that filing now carries maximum deterrent weight — I genuinely don't know. But it is not accidental that the lawsuit arrived while the IPO clock is running.
The broader shift I see underneath all of this is a transition from the era of scarcity to the era of abundance — and all the messy renegotiations that come with it. When there were only one or two credible frontier models, the platforms needed them badly enough to accept almost any terms. Now there are six or seven. GPT-5.6, Claude Fable 5, Grok 4.5, Gemini, Muse Spark. The price per token has collapsed. The performance gaps between the top models are narrowing. The platforms have options. And when you have options, you renegotiate.
What I don't know — what nobody knows yet — is what the new equilibrium looks like. The cooperation era had a kind of logic to it: everyone needed everyone else, so the deals got done. The competition era has a different logic, and it's sharper-edged. Lawsuits. Distribution switches. IPO pressure. The models that get embedded in the most surfaces win the most, but the surfaces are now negotiating back.
There is something almost ecological about it. A period of symbiosis giving way to something more competitive as the resources become abundant enough to fight over. The organisms involved are trillion-dollar companies, so the stakes are not small. But the underlying dynamic is not unfamiliar.
The cooperation phase is over. What comes next is harder to predict — and probably more interesting to watch.