Business IEA Says Data Center Power Demand Could Double to 1,000 TWh by 2030, With AI Supplying Three Quarters of the Growth Chatbots ChatGPT Now Transcribes as You Talk: OpenAI Removes the 10-Minute Cap and Streams Words Live on Every Plan Business Ex-Alibaba Executive's Everful Raises $48.5 Million to Put AI Agents in Charge of Sourcing From 30,000 Factories AI Agents Google Built a Mirror Dimension of the Internet to Train Its Chrome AI Agents Without Breaking the Real Web Productivity Canva Finally Adds an Off Switch for Its Generative AI Features, and a Filter to Hide AI-Made Templates Business Anthropic Begins Trading Today: 980 Million Shares at $83, a $1.5 Trillion Valuation and the Largest IPO on Record News Google Books Blue Origin New Glenn to Launch TPU-Carrying AI Satellites in Early 2029 Security Hackers Used the AI Penetration Tool ARTEX to Disrupt European Commission Systems for Over Ten Days Business IEA Says Data Center Power Demand Could Double to 1,000 TWh by 2030, With AI Supplying Three Quarters of the Growth Chatbots ChatGPT Now Transcribes as You Talk: OpenAI Removes the 10-Minute Cap and Streams Words Live on Every Plan Business Ex-Alibaba Executive's Everful Raises $48.5 Million to Put AI Agents in Charge of Sourcing From 30,000 Factories AI Agents Google Built a Mirror Dimension of the Internet to Train Its Chrome AI Agents Without Breaking the Real Web Productivity Canva Finally Adds an Off Switch for Its Generative AI Features, and a Filter to Hide AI-Made Templates Business Anthropic Begins Trading Today: 980 Million Shares at $83, a $1.5 Trillion Valuation and the Largest IPO on Record News Google Books Blue Origin New Glenn to Launch TPU-Carrying AI Satellites in Early 2029 Security Hackers Used the AI Penetration Tool ARTEX to Disrupt European Commission Systems for Over Ten Days

Apple Intelligence: What Runs on the Device and What Does Not

Apple Intelligence: What Runs on the Device and What Does Not

Apple splits AI work between the phone and private cloud infrastructure. The split determines both privacy and capability.

Apple's approach to AI is defined by where the computation happens. Small models handle tasks locally; larger requests route to servers Apple operates with a stated privacy design.

On-device tasks

  • Text summarisation and rewriting in system apps.
  • Notification prioritisation and smart replies.
  • Local semantic search across photos and messages.
  • Image cleanup and basic generation on supported hardware.

Server-assisted tasks

Requests requiring a larger model are sent to Apple's infrastructure, where the company states that inputs are not retained and are not accessible to Apple staff. Independent verification of those claims is limited, which is a fair criticism even from supporters.

What this means for users

Hardware generation determines which features are available at all, and feature availability by language and region remains uneven. For privacy-sensitive work, local features are the safer default — but they are also the least capable.

Comments (0)

Log in to join the discussion

Log In

No comments yet