Signal

Court docs: OpenAI says Apple's ChatGPT integration to power Apple Intelligence on iPhones "dramatically underperformed" after it launched in 2024

First reported by Ft ·

The signal ●●●● Compiled by AI from Ft, Techmeme, PYMNTS, Kirkville, iPhone in Canada and 1 more
Why you might care

If you use AI chatbots, the actual performance of OpenAI's Apple integration is now a point of legal contention.

What happened

OpenAI has filed a motion seeking dismissal of Elon Musk's antitrust lawsuit, citing SpaceX's public securities filings as evidence. The lawsuit, filed by Musk's companies, alleged that Apple's integration of ChatGPT into Apple Intelligence provided OpenAI an unfair competitive advantage. OpenAI argues that SpaceX's disclosures to investors present a different competitive landscape than alleged in the complaint, thereby undermining claims of commercial harm. Furthermore, OpenAI contends that the integration with Apple generated fewer new users than projected, and the performance was "dramatically underperformed." This reduced user growth, they argue, contradicts the assertion of market dominance. The case is proceeding against OpenAI after Musk's X Corp. and SpaceXAI settled their claims with Apple.

What it means

OpenAI's defense hinges on portraying the Apple partnership's underwhelming user adoption as evidence against claims of market monopolization, suggesting the deal did not confer the anticipated significant competitive edge. This argument aims to debunk the core of Musk's allegations that the integration unfairly bolstered OpenAI's market position among iPhone users.

The legal battle highlights the challenges in quantifying the impact of AI integrations and the potential for underperformance to become a key defense strategy in antitrust cases. As OpenAI seeks to dismiss the suit before trial, the court's consideration of these performance metrics could set a precedent for future disputes over AI market competition and the definition of unfair advantage.

AI-written summary. May contain errors.