OpenAI Just Called Out Apple in Court — And It Changes Everything

The Public Unraveling of Apple’s Biggest AI Bet

OpenAI didn’t just file a legal brief against Elon Musk’s xAI — it filed a public indictment of Apple. The court documents, unsealed in an antitrust dispute that has surprisingly little to do with Apple’s actual behavior, contain language that reads like a disappointed partner’s resignation letter. The integration of ChatGPT into Siri, launched in December 2024 as one of Apple’s most高调 AI announcements, was, according to OpenAI’s own filings, “dramatically underperforming” and “persistently underperforming.”

The detail that matters most is not the legal strategy. It’s the timeline. By January 2025 — roughly five weeks after the feature went live — OpenAI had already cut its forecast for incremental logged-in weekly active users from the partnership. That is an unusually fast retreat from a deal that both companies marketed as foundational.

A Non-Exclusive Deal That Felt Exclusive

Here is what the filing actually proves, stripped of redactions: Apple never promised OpenAI exclusivity. The contract explicitly reserved Apple’s right to integrate competing services. Apple told OpenAI it planned to start with one provider and add more. It said the same thing publicly when announcing the feature.

Then Apple signed a deal with Google. Its newest models — the ones powering Apple Intelligence beyond the ChatGPT extension — are built on Gemini.

xAI’s lawsuit alleged an exclusive deal harmed its growth. OpenAI’s response was essentially: the contract says no, and even if it did, the damage is negligible. Dr. Catherine Tucker, OpenAI’s expert witness, calculated that the share of generative AI consumers accessing ChatGPT through Apple Intelligence was de minimis across every metric. In plain language: very few people used it, so it couldn’t have foreclosed competition.

Musk’s companies dropped their claims against Apple earlier this month. OpenAI remains the sole defendant. The trial is set for January 2027.

What This Reveals About the Power Shift

The real story here is not the lawsuit. It is the reversal of roles. For years, the dynamic was clear: platform owners like Apple controlled access to device-level AI experiences, and model providers begged for integration. The iPhone was the gate. You got in on Apple’s terms, or you didn’t.

That dynamic is inverting. OpenAI is now the party that can afford to be public about a partnership failing — because it has alternatives. Google. Its own apps. The web. Apple, by contrast, has staked a visible claim on AI being the next Siri moment, and the numbers are embarrassing.

The evidence accumulates. Apple’s $250 million class-action settlement over Siri’s delayed AI launch — roughly $25 per eligible iPhone user — signals that even Apple’s internal timelines were slipping. iOS 27’s new Siri features are behind a waitlist. Code uncovered by researchers shows Apple designed its Siri architecture to accept third-party models like Claude through a mechanism called “Model Delegation,” which means the ChatGPT extension was always meant to be swappable — and apparently, it was swapped out in practice for Gemini-powered capabilities.

The Second-Order Implications

If this pattern holds, the app ecosystem model that Apple has defended for two decades is vulnerable in a new way. The old argument was: developers need iOS because iPhone users spend money. The emerging argument is: AI model providers need distribution, but they no longer need Apple specifically. Google has Gemini. Microsoft has Copilot baked into Windows and Office. Amazon has Claude through its AWS infrastructure. Meta has its own stack. The gate is not closed.

For Apple, the risk is sharper. Its hardware advantage depends on software experiences that feel uniquely good. AI is supposed to be the thing that makes that argument again. If the ChatGPT integration is a footnote and the real intelligence comes from Gemini via a deal that wasn’t exclusive, then Apple is paying premium prices for second-tier AI access while its competitors are building first-party models into their platforms.

The OpenAI filing also reframes the antitrust debate. xAI’s claim was that Apple and OpenAI locked out competitors. OpenAI’s counter is that the lockout never existed and the impact was zero. Both sides are wrong about the deeper question: whether platform control over AI integration will matter less as model providers build direct channels to users.

Who Wins, Who Loses

OpenAI wins credibility. It enters the courtroom as the honest broker rather than the monopolist, which matters when the case is about market power. Google wins because its models now underpin Apple’s core AI experience without attribution. Apple loses the narrative about AI being a differentiator.

xAI loses a defendant but arguably wins the argument — even though its own claims against Apple were dropped, the record now confirms that any exclusivity effect was nonexistent. Claude’s potential entry into Siri, if the code is any indication, adds another layer to Apple’s strategy of keeping options open rather than committing to one provider.

What Happens Next

The January 2027 trial will likely focus on whether OpenAI’s market power constitutes monopoly exploitation — a question the Apple integration evidence may actually hurt OpenAI’s case, since the data shows minimal user uptake. But the public record is already set. Every future partnership Apple announces will be measured against this moment: a company admitting on the record that its biggest AI integration was a disappointment from the start.

The power shift is real. Model providers are no longer supplicants. Platform owners are learning that integration is not strategy — and that the companies controlling the models may soon control the relationship with the user.