On September 24, 2026, the Financial Times reported that newly unsealed court filings have revealed, for the first time, the inside story of the breakdown of the OpenAI–Apple partnership. Notably, the disclosure was not dug up by reporters — it was written by OpenAI itself in legal filings. On the once high-profile alliance, OpenAI stated in the documents that by the summer of 2025, "it was clear that Apple's integration of ChatGPT was performing far below expectations."
According to the FT's account of the filings, the two companies reached an agreement in June 2024 for ChatGPT to power Apple Intelligence. OpenAI had expected Apple's promotion and brand "halo effect" to translate into subscription growth. Instead, the feature got off to a "slow start" within a month of launch, and OpenAI cut its weekly active user forecasts. From the moment those internal projections were revised down, the partnership's trajectory was essentially set.
Timeline: from grand opening to going separate ways
Lining up the key moments of this sub-three-year partnership reads like a chronicle of dashed expectations. In June 2024, the deal was unveiled with fanfare at WWDC, with ChatGPT integrated into Apple Intelligence; OpenAI was banking on Apple's reach and brand endorsement. User growth then fell short and forecasts were cut. In January 2026, Apple switched to Google, using Gemini models to power a rebuilt Siri AI. In May 2025, Bloomberg reported cracks in the relationship, with OpenAI hiring outside counsel to evaluate a breach-of-contract suit. In July 2026, Apple sued OpenAI over trade secrets, alleging it poached former Apple employees including Tang Tan, now OpenAI's hardware chief. From a "kingmaker" narrative to courtroom adversaries — the honeymoon was far shorter than outsiders realized.
Breaking down the dispute: why OpenAI itself released this material
These filings were not voluntary PR from OpenAI. They come from the antitrust lawsuit Elon Musk's xAI filed in August 2025, which alleged that the exclusive Apple–OpenAI arrangement squeezed out rivals like Grok. Apple settled with xAI last week, and OpenAI has filed a legal motion demanding disclosure of the settlement terms, which brought the documents into public view.
OpenAI's argument in the filings is strikingly blunt: the partnership did not unfairly boost its market position — on the contrary, its market share was broadly declining during the same period, while Google, Anthropic, Meta, and Grok were winning users. In other words, OpenAI is countering the antitrust claims in court with "this deal didn't actually help us much"; the "far below expectations" language is both a status report and litigation strategy. Apple has not publicly responded to these claims.
While the old partnership falls apart, OpenAI keeps hunting for new distribution channels — from enterprise agents to system-level integrations. The compliance controversies around moves like the OpenAI agent that entered Australia's Medicare system show that OpenAI's distribution push keeps running into real-world friction.
Impact: wins and losses on all four sides
For OpenAI, the setback is direct: it has lost one of the best mobile distribution channels on the planet, and its most promising subscription growth engine. That Apple's promotion across billions of active devices failed to convert into subscriptions should worry OpenAI more than the breakup itself.
For Apple, the pivot was nearly inevitable. Its January 2026 switch to Gemini for the rebuilt Siri AI amounted to declaring a "multi-model strategy" — no more putting all eggs in one basket, and no more exclusivity for a single partner.
For xAI and Grok, the lawsuit has partly achieved its goal. Apple's settlement with xAI last week loosened the exclusivity barrier; the very allegation of squeezing out rivals has itself become pressure on exclusionary arrangements.
For the industry, the fight over AI distribution channels enters a new phase: handset makers are plugging in multiple models while keeping in-house options open, and the single "default AI" slot is getting harder for any one player to monopolize.
Our take
First, the "kingmaker" narrative needs correcting. Apple was seen as a traffic behemoth that could turn anything it touched into gold — get into the iPhone and you've won. This partnership proves that distribution does not equal conversion: users don't pay for pre-installation, they pay for the experience. Subscription growth for AI products comes from the product itself, not the channel's halo.
Second, the exclusive-partnership model is ending. From this breakup to Apple's switch to Gemini, the direction is clear: platform owners want models that are swappable, comparable, and replaceable at any time; for model companies, an exclusive deal may be a shackle rather than a moat.
Third, this is arguably good news for ordinary users. With Apple moving to a multi-model strategy, iPhone users may eventually get to choose their model inside Siri — the exclusive seat OpenAI lost could become the very place where users gain choice.
One final note on sourcing: everything above about the partnership's inside story is either "OpenAI's account in court filings" or "as reported by the Financial Times" — not facts independently confirmed by reporters; Apple has not publicly responded, and the other side of the story has yet to be told.
