
Credit: Unsplash/CC0 Public Domain
Apple Inc. is considering using human PBC or Openai’s artificial intelligence technology to power new versions of Siri, and is bystanding its own in-house model with a potential blockbuster move aimed at its burning AI efforts.
According to people familiar with the discussion, iPhone makers have spoken with both companies about the use of SIRI’s large-scale language models. It asked them to train versions of models that could run on Apple’s cloud infrastructure for testing, said those who discussed personal deliberations and asked not to be identified.
If Apple finally moves forward, it represents a monumental reversal. The company is currently enhancing most of its AI capabilities, calling the Apple Foundation model and planning a new version of its voice assistant to run its technology in 2026.
The switch to Anthropic’s Claude or Openai’s Siri’s ChatGPT model acknowledges the company is struggling to compete in Generic AI, the most important new technology of decades. Apple already allows ChatGpt to answer Siri’s web-based search queries, but the assistant itself is powered by Apple.
People said the investigation into Apple’s third-party models is early on, and the company has not made a final decision on using them. A competing project called LLM Siri, which uses an in-house model, remains active in development.
Based in Cupertino, California, you can allow Apple to provide SIRI functionality on Android phones on comparable AI assistants.
Representatives from Apple, Anthropic and Openai declined to comment. Apple’s shares closed more than 2% after Bloomberg reported on the deliberations.
Siri is struggling
The project to evaluate the external model was launched by Siri Chief Mike Rockwell and head of software engineering Craig Federighi. They were given Siri’s supervision after the duties were removed from the direction of the company’s AI chief John Giannandorea. He stood by in the wake of lukewarm waters against delays in Apple Intelligence and Siri features.
Rockwell, who previously launched the Vision Pro headset, took on the role of SIRI Engineering in March. After taking over, he instructed the new group to assess whether Siri would use Apple’s AI models or third-party technologies such as Claude, ChatGpt, and Google Gemini from Alphabet Inc. to handle queries for a better job.
After multiple tests, people said Rockwell and other executives concluded that human technology was the most promising for Siri’s needs. That’s why Adrian Perica, the company’s vice president of corporate development, will begin debating with humanity about Claude’s use, people said.
The SIRI Assistant was released in 2011 with Origie, but has lagged behind the popular AI chatbots, and Apple’s attempts to upgrade its software are hampered by engineering snags and delays.
A year ago, Apple announced a new Siri feature that utilizes user personal data to analyze content on-screen to better fulfill queries. The company has also demonstrated technology that allows Siri to have more accurate control over apps and features across Apple devices.
Enhancements are not ready. Apple initially announced plans for an early release in 2025, but ultimately delayed the launch indefinitely. They are planned for next spring, as reported by Bloomberg News.
AI uncertainty
People with knowledge of Apple’s AI team say it works with a high degree of uncertainty and lack of clarity. Apple has already approved a multi-billion dollar budget for 2026 to run its own model through the cloud, but plans beyond that are phenomenal.
Still, Federighi, Rockwell and other executives are increasingly open to the idea that embracing external technology is the key to a short-term shift. They don’t think the need for Apple to resort to their own model (which they consider to be inferior now) if people can partner with third parties instead.
The licensed third-party AI reflects the approach taken by Samsung Electronics Co. The company brands features under the Galaxy AI umbrella, but many of those features are actually based on Gemini. Humanity is already used in that part by Amazon.com Inc. to provide the power of the new Alexa+.
In the future, if unique technologies improve, executives believe that Apple should own the AI model, given that it is becoming more important to the operation of its products. The company is working on a range of projects, including tabletop robots and glasses that use AI richly.
Apple has recently been considering gaining confusion to bolster its AI job, Bloomberg reports. We also had a brief discussion with Thinking Machines Lab, an AI startup founded by former Openai Chief Technology Officer Mira Murati.
Healing morale
The Apple model is developed by a team of about 100 people run by Ruoming Pang, a well-known Apple engineer who joined from Google in 2021 and led the work. He reports to Daphne Luong, senior director of AI research.
Luong is one of the top Lis at Giannandrea, and the Foundation Models team is one of the few key AI groups that still report to Giannandrea. Even in that area, Federigi and Rockwell have played a bigger role.
Regardless of the path it took, the proposed shift weighed the team. This includes some of the most sought-after talent in the AI industry.
Some members are internally signaling that the company is unhappy with its consideration of third-party technology, and at least in part creates perceptions of criticising the company’s AI shortcomings. They say they can leave for the multi-million dollar package floating by Meta Platforms Inc. and Openai.
Facebook and Instagram owner Meta offers annual pay packages to $10 million to $40 million or even more engineers to join the new Superintelligence Labs group, according to people with knowledge of the issue. Apple is known to be able to load half or even fewer into the open market, often with AI engineers.
Tom Gunter, one of Apple’s most advanced and leading language model researchers, left last week. He worked for Apple for about eight years, but some colleagues find it difficult to trade him, given his unique skill set and his Apple competitors willing to pay exponentially for their talent.
Again, Apple has lost most of the team behind MLX, the leading open source system for developing machine learning models on the latest Apple chips. After the engineers threatened to leave, Apple did an anti-off to keep them.
Humanity’s debate
In discussions with both Anthropic and Openai, the iPhone makers requested custom versions of Claude and ChatGpt that can run on Apple’s private cloud computing servers.
Rather than relying on third-party infrastructure, Apple believes running models on its own chips housed in Apple-controlled cloud servers would better protect user privacy. The company has already internally tested the feasibility of the idea.
Other Apple Intelligence features include AI models that exist on consumer devices. These models are less luxurious and powerful than the cloud-based version, but are used for tasks such as short email summaries and writing Genmojis.
Later this year, Apple will open models on the device to third-party developers, allowing app makers to create AI capabilities based on that technology.
The company has not announced plans to allow apps to access cloud models. One reason for this is that cloud servers still don’t have the ability to handle flooding of new third-party features.
The company is currently not working on moving away from its internal model for device or developer use cases. Still, it is fear among the engineers on the foundation modeling team that they can move to third parties for SIRI.
Last year, Openai offered to train models on Apple’s devices, but the iPhone maker was not interested.
Since December 2024, Apple has been using Openai to handle some features. In addition to responding to SIRI’s world knowledge questions, ChatGPT allows you to write blocks of text in the writing tool functionality. Later this year, iOS 26 will have the ChatGPT option for image generation and on-screen image analysis.
People say that while discussing potential arrangements, Apple and Anthropic are in agreement on preliminary financial terms. AI startups are seeking annual fees of billions of dollars, which are rapidly increasing every year. With the struggle to reach the deal, they said Apple is considering working with Openai and others if it goes ahead with third-party plans.
Management Shift
If Apple signs an agreement, the impact of Giannandrea, who joined Apple from Google in 2018 and advocate for large-scale language model development within the company, will continue to shrink.
In addition to losing Siri, Giannandrea was deprived of responsibility for Apple’s Robotics unit. Additionally, in a previously unreported move, the company’s core ML and app intent team (the group responsible for frameworks that allow developers to integrate AI into apps) has moved to Federighi’s software engineering organization.
Apple’s Foundation Models team was building a large language model that allowed employees and external developers to write code in Xcode, the programming software. The company killed a project announced last year as prompt support a month ago.
Instead, Apple is rolling out a new Xcode later this year that will allow you to take advantage of third-party programming models. App developers can choose from ChatGpt or Claude.
2025 Bloomberg LP Tribune Content Agency, distributed by LLC.
Quote: Apple weighs using siri in a massive reversal using humanity or Openai (July 1, 2025) Retrieved from July 1, 2025 https://techxplore.com/news/2025-07-Apple-antropic-openai-power-siri.html
This document is subject to copyright. Apart from fair transactions for private research or research purposes, there is no part that is reproduced without written permission. Content is provided with information only.
