OpenAI’s partnership with Apple, announced to considerable fanfare in June 2024, is fracturing. The AI company’s lawyers are actively working with an outside legal firm on options that could include sending Apple a formal notice alleging breach of contract, according to Bloomberg, which cited people familiar with the deliberations. No lawsuit has been filed, and OpenAI says it still hopes to resolve the dispute without going to court. But the company has concluded that Apple failed to hold up its end of a deal that was supposed to turn ChatGPT into a default feature of the world’s most valuable consumer ecosystem.

The core grievance is distribution. OpenAI believed that weaving ChatGPT into Siri, Writing Tools, and Image Playground would drive a significant number of iPhone users toward paid subscriptions – potentially generating billions of dollars per year, according to people familiar with the company’s expectations. Instead, Apple’s implementation buried the integration behind friction. Users must explicitly invoke the word “ChatGPT” when speaking to Siri to trigger OpenAI’s models. Responses appear in constrained windows with limited information. Apple customers overwhelmingly prefer using the standalone ChatGPT app, according to user studies conducted by OpenAI.

What went wrong

The deal, struck when Apple was scrambling to catch up on generative AI, gave iPhone users the ability to access ChatGPT results through Siri, generate text, analyse images via Visual Intelligence, and create pictures in Image Playground. Apple took a cut of subscription revenue generated through its platforms. OpenAI got what it believed would be prime placement inside an ecosystem of more than a billion active devices.

That placement never materialised in the way OpenAI expected. An executive at the company, speaking to Bloomberg on condition of anonymity, said Apple had not made an honest effort to promote the integration. OpenAI now believes the implementation has actively damaged its brand, with the limited, windowed responses creating an impression of inferior capability compared with the full ChatGPT experience.

During initial discussions in 2024, Apple characterised the opportunity as comparable to its search deal with Google in Safari – a partnership that generates tens of billions of dollars annually for both sides. That comparison has proved spectacularly inapt. The same executive said OpenAI had been told to take a leap of faith, and that the leap had not paid off.

A relationship under pressure from all sides

The rift with Apple is not happening in isolation. OpenAI is simultaneously fighting a federal trial with Elon Musk over its nonprofit-to-profit conversion, with potential damages of $150 billion. It recently renegotiated its exclusive deal with Microsoft, capping revenue-sharing payments at $38 billion through 2030 and shifting to a non-exclusive licensing model that allows OpenAI to serve customers on any cloud provider. Amazon, meanwhile, has deepened its investment in Anthropic with an additional $5 billion.

Apple has its own list of grievances. The company has been concerned about OpenAI’s privacy practices. And OpenAI’s $6.5 billion acquisition of io, the device startup co-founded by former Apple design chief Jony Ive, has created a direct competitive threat. The business, now run by former Apple executives Tang Tan and Evans Hankey, is working to build an alternative to the iPhone. Apple executives have been, as Bloomberg put it, fuming over OpenAI’s recruitment of Apple hardware engineers, with the AI company offering stock packages worth millions more than Apple provides.

Any legal action by OpenAI is unlikely before the conclusion of the Musk trial, according to Bloomberg’s sources. The timing suggests OpenAI is managing its legal exposure carefully, avoiding a second front while the first remains unresolved.

Apple’s pivot away from OpenAI

The partnership’s deterioration is happening as Apple prepares to diminish OpenAI’s role in its software. iOS 27, expected to be unveiled at the Worldwide Developers Conference on 8 June, will introduce a system called Extensions that allows users to install rival AI chatbots from the App Store and route Siri queries, writing tasks, and image generation through whichever model they choose. Apple is testing integrations with both Anthropic’s Claude and Google’s Gemini.

Separately, Apple struck a deal late last year to use Google’s Gemini models as the foundation for its own AI capabilities, paying roughly $1 billion annually for a custom 1.2-trillion-parameter model that will power the next generation of Siri. OpenAI was considered for this deeper integration but declined to participate, feeling burned by the original relationship.

The OpenAI executive told Bloomberg that Apple’s embrace of other providers is not driving the legal dispute, since the partnership was never meant to be exclusive. The new Extensions system might even benefit ChatGPT by giving it more prominent placement through a model-picker interface. But the broader trajectory is clear: Apple is building an AI architecture designed to reduce its dependence on any single provider, and OpenAI is moving from privileged partner to one option among several.

The pattern

Apple’s AI strategy has been defined by a series of forced compromises. The company marketed AI features for the iPhone 16 that were not ready, leading to a $250 million class-action settlement over false advertising. It partnered with OpenAI because its own models were inadequate, then found the partnership unsatisfying. It turned to Google for the underlying intelligence that its in-house team could not deliver. At each stage, the company that built its reputation on vertical integration has been forced to depend on others for the capability its customers expect.

For OpenAI, the Apple experience is a lesson in the limits of distribution without control. Being inside the iPhone sounded like a growth engine. In practice, it meant accepting Apple’s design choices, Apple’s revenue terms, and Apple’s willingness – or unwillingness – to promote the product. The company that believed it was getting a Safari-scale partnership got something closer to a buried settings toggle.

Whether the dispute escalates into litigation or resolves through renegotiation, the strategic picture has already shifted. OpenAI is building hardware. Apple is building its own AI stack. The deal that once symbolised their mutual need now illustrates why the two largest ambitions in consumer technology – owning the device and owning the intelligence  may be fundamentally incompatible as long-term partnerships.