Apple Considers OpenAI, Anthropic for Siri AI

The Cupertino company is reportedly testing ChatGPT and Claude to improve Siri: according to Bloomberg, Apple is not satisfied with its own proprietary models and is looking for external solutions to fill the gap with the competition

Apple is still struggling to relaunch Siri with advanced capabilities based on generative artificial intelligence, and to do so it is reportedly considering an option that was previously unthinkable: relying on external companies. According to Bloomberg's Mark Gurman, the company led by Tim Cook has started tests with advanced linguistic models (LLM) from OpenAI (ChatGPT) and Anthropic (Claude), evaluating their performance on simple requests within its private cloud environments.

The “LLM Siri” project, the evolution of the AI-powered voice assistant, has suffered a series of delays, including the postponement of the launch initially scheduled for March. The lack of convincing results has led Apple to an internal restructuring: Mike Rockwell, previously head of the Vision Pro team, has been appointed head of the AI ​​and Siri division after CEO Tim Cook himself reportedly lost confidence in the previous leader of the project, John Giannandrea.

The language model comparison isn’t limited to internal developments: In addition to testing Claude and ChatGPT, the Rockwell team also looked at Gemini, Google’s model already integrated into Android devices and the Pixel line. But it’s Anthropic’s Claude that has yielded the most promising results so far, according to Bloomberg. Apple’s interest in Claude comes amid a competitive landscape in which Google has struck deals with Samsung and Motorola to integrate its AI models, while Apple has also reportedly explored the possibility of acquiring Perplexity, a fast-growing startup in the conversational AI sector.

A significant detail: during the WWDC keynote in June, the new Siri with artificial intelligence was never shown. A significant absence, explained by Greg Joswiak himself (SVP of global marketing at Apple), who admitted that the technology “has not yet reached our quality standards”. Despite tests with external partners, Apple has not completely abandoned the development of its own linguistic models. The hypothesis of integrating a third-party LLM into Siri is still in a preliminary phase and nothing has been decided. But the picture is clear: Apple is moving cautiously, aware of being late in the race for generative artificial intelligence, but determined to fill the gap. And this time, it could also do so by asking for help from the competition.