Apple (AAPl) is in early talks to use Google’s AI model Gemini to power a revamped version of its voice assistant Siri, Bloomberg News reported, citing people with knowledge of the matter.
Apple recently approached the Alphabet (NASDAQ:GOOG) (NASDAQ:GOOGL) unit Google to explore building a custom AI model that would serve as the foundation of the new Siri next year, the report added.
Apple and Google did not immediately respond to a request for comment from Seeking Alpha.
The iPhone maker has been trying to catch up in the AI race. Earlier this year, Apple also explored collaborations with Anthropic PBC and OpenAI, weighing if Claude or ChatGPT could power Siri, the report noted.
Apple is still several weeks away from deciding if it would continue using internal models or move to a partner for Siri. The company has also not decided who that partner could be, the report added.
The potential pivot follows delays to a long-awaited upgrade of Siri, which would fulfill commands by tapping into personal data and let users navigate devices through voice. The Siri update was slated for this past spring but was postponed by a year because of engineering setbacks, the report noted.
The failure led Apple to sideline AI chief John Giannandrea from Siri development. The project is now being helmed by software head Craig Federighi and Vision Pro headset creator Mike Rockwell, who are mulling the use of outside help as a potential path forward, the report noted.
The Siri update was initially designed around technology developed by the Apple Foundation Models team. This group also developed the on-device large language models that power Apple Intelligence features such as summarizing text and creating custom emoji.
Trying to fix Siri’s flaws and bring the delayed features to market, Federighi, Rockwell and Apple’s corporate development team, headed by Adrian Perica, began meeting with Anthropic and OpenAI about a potential deal.
Apple is simultaneously developing two versions of the new Siri — one known as Linwood that is powered by its models and another dubbed as Glenwood, which runs on outside technology, the report added.
Executives had long seen Anthropic as the leading option for a collaboration, but the financial terms set by that company led Apple to broaden the search and bring others to the table. Apple has not ruled out sticking to its own models, the report noted.
These discussions are separate from other agreements to integrate chatbots into Apple Intelligence. Last year, Apple added ChatGPT as a Siri fallback for general knowledge questions. And both Apple and Google have already publicly indicated plans for a similar Gemini integration, the report added.
The discussions about using Gemini models to power Siri remain exploratory, with no formal commercial negotiations currently underway, the report added. Google has made similar deals previously and powers much of the AI functionality on phones sold by Samsung Electronics (OTCPK:SSNLF).
Meanwhile, Apple’s AI models teams have continued to see changes. In July, Ruoming Pang, the team’s chief architect, left for Meta Platforms (META). He was offered a $200M package and a senior role in Meta’s new Superintelligence Labs unit, as per the report.
Earlier on Friday it was reported that Frank Chu, who led Apple AI teams for cloud infrastructure, training, and search, would join Meta’s Superintelligence Labs to work on AI infrastructure.
Some software engineering leaders have pitched a related idea of replacing the models used for non-Siri AI features on Apple devices. The company has generally tried to maintain ownership of the AI features which run on consumers’ systems, aiming to ensure security and privacy.
Regarding a Siri partnership, third-party models would run on Apple’s Private Cloud Compute servers that use Mac chips for remote AI processing. This means the external Siri models would not run on devices themselves, the report added.
During a quarterly earnings call, CEO Tim Cook declined to comment on the use of third-party models, noting that any discussion would reveal the company’s plans.
Apple has already begun enlisting partners for some consumer features and its internal operations. In the iOS 26 operating system, Apple is offering a ChatGPT option for generating images, and it ended a publicly announced project to develop its own generative AI-based coding system. Instead, Apple is using ChatGPT and Claude, the report added.
However, work within Apple’s Foundation Models team, continues. The company recently started testing its first trillion-parameter model, a major pivot from the 150 billion systems currently running in Apple’s AI data centers, the report noted.
Currently, Apple has no plans to offer this more powerful system to customers. For now at least, it will be only used for research, the report added.