Apple says its AI is still private, even when it's running on Google's servers

28d ago · US · primary source: arstechnica.com

Apple confirmed at its Worldwide Developers Conference that its new Siri AI assistant runs on Google servers using Nvidia hardware, while the company maintains that user data remains private through encryption and on-device processing [1]. The arrangement marks a departure from Apple's prior AI infrastructure, which relied entirely on local device processing or its own Private Cloud Compute system [1]. Apple executives, including software chief Craig Federighi, briefed press after the keynote to detail how the company intends to preserve privacy despite routing requests through Google's data centers [1]. The company has long positioned privacy as a competitive advantage, using encryption designed to block access by anyone — including Apple employees — and emphasizing on-device processing to minimize data leaving a user's device [1]. Apple's shift to Google's Gemini models reflects the computational demands of modern generative AI [1]. The language and reasoning models that can operate locally on an iPhone or Mac are relatively small, limiting their capabilities [1]. Gemini, first announced by Google in December 2023, is a family of large language models trained natively on multiple data types including text, code, images, audio, and video [2]. The technology is distributed in tiers, from efficient on-device versions to high-compute models for complex reasoning [2]. Apple's Private Cloud Compute system was a partial solution but required Apple's own server hardware; supporting Siri AI at scale would have required a data center buildout the company has so far avoided [1]. The partnership places two of the Big Five U.S. technology companies — Apple and Alphabet, Google's parent — in a direct operational dependency [10]. Generative AI tools have proliferated since the AI boom of the 2020s, driven by advances in deep neural networks and transformer-based large language models [3]. The sector has drawn scrutiny over data practices, energy consumption, and the use of copyrighted training material [3]. Apple, founded in 1976 and headquartered in Cupertino, California, has grown to become the world's largest company by market capitalization, valued at over $4 trillion as of October 2025 [4]. The company has consistently ranked among the world's most valuable brands [4]. Apple says its AI models remain encrypted even when processed on Google's infrastructure, extending its privacy architecture to third-party servers [1]. The company did not disclose the specific Gemini model tier being used, nor the financial terms of the arrangement [1].

infrastructure

Background sources we checked (10)
  • en.wikipedia.org ↗ Gemini (also known as Google Gemini and formerly known as Bard) is a generative artificial intelligence chatbot and virtual assistant developed by Google. It is powered by the family of large language models (LLMs) of the same name, after previously being based on LaMDA and PaLM …
  • en.wikipedia.org ↗ Generative artificial intelligence (GenAI) is a subfield of artificial intelligence (AI) that uses generative models to generate text, images, videos, audio, software code (vibe coding) or other forms of data. These models learn the underlying patterns and structures of their tra…
  • en.wikipedia.org ↗ Apple Inc. is an American multinational technology company headquartered in Cupertino, California, in Silicon Valley, and known for consumer electronics, software and online services. Founded in 1976 as Apple Computer Company by Steve Jobs, Steve Wozniak and Ronald Wayne, the com…
  • arxiv.org ↗ Tag-based tracking ecosystems help users locate lost items, but can be leveraged for unwanted tracking and stalking. Existing protocol-driven defenses and prior academic solutions largely assume stable identifiers or predictable beaconing. However, identifier-based defenses funda…
  • arxiv.org ↗ Online services often require users to agree to lengthy and obscure Terms of Service (ToS), leading to information asymmetry and legal risks. This paper proposes TOSense-a Chrome extension that allows users to ask questions about ToS in natural language and get concise answers in…
  • arxiv.org ↗ Predicting stock market movements remains a persistent challenge due to the inherently volatile, non-linear, and stochastic nature of financial time series data. This paper introduces a deep learning-based framework employing Long Short-Term Memory (LSTM) networks to forecast the…
  • arxiv.org ↗ In this paper, we apply quantum machine learning (QML) to predict the stock prices of multiple assets using a contextual quantum neural network. Our approach captures recent trends to predict future stock price distributions, moving beyond traditional models that focus on entire …
  • arxiv.org ↗ Asymmetric Numeral Systems (ANS) proposed by Jarek Duda are high-performance distortionless data compression schemes that can achieve almost the same compression performance as arithmetic codes with less arithmetic operations than arithmetic coding. The ANS is widely used in vari…
  • en.wikipedia.org ↗ Big Tech, also known as the tech giants or tech titans, are the largest and most influential technology companies in the world. The term Big Tech commonly refers to the five U.S. technology companies Microsoft, Apple, Alphabet, Amazon, and Meta, also known as the Big Five. The Bi…
  • en.wikipedia.org ↗ Apple Maps is a web mapping service developed by Apple. As the default map system of iOS, iPadOS, macOS, tvOS, visionOS, and watchOS, it provides directions and estimated times of arrival for driving, walking, cycling, and public transportation navigation. A "Flyover" mode shows …

Sources

Spot something wrong? Report an issue