Urgent.News

What's breaking now, across thousands of outlets.

AI

Künstliche Intelligenz: Apple plant Datenwarnung bei KI-Agenten auf Mac-Computern

Sogenannte KI-Agenten sind auf weitreichenden Zugang zu Nutzerdaten angewiesen. Apples Mac-Computer bieten diese Möglichkeit - anders als andere Geräte der Firma. Jetzt sollen Nutzer gewarnt werden.

Künstliche Intelligenz: Apple plant Datenwarnung bei KI-Agenten auf Mac-Computern

Apple is taking steps to inform Mac users about the risks associated with independently operating AI agents before granting such software extensive access to their data. Some developers have previously abused access capabilities that could expose all data on the devices, prompting Apple to ensure users are aware of the dangers before agreeing to data access for AI.

AI agents are designed to autonomously perform tasks for users, and according to AI companies, their new agents can now perform almost any task a computer can do. To be truly useful in everyday life, they require deep access to the user's information, including emails, chats, calendars, and potentially even bank accounts. Apple's Macs have a feature called "Full Disk Access" which grants extensive data access on the hard drive, initially implemented for backup programs to create smooth copies of the computer.

However, AI software developers have also exploited this feature. As AI agents become more capable and autonomous, the risks associated with them are also expected to increase. While iPhones and iPads have much more restricted data access options compared to Macs, Apple is working to make its in-house AI assistant, Siri, more useful by leveraging AI functions that benefit from user data on the devices.

Apple emphasizes that data is processed solely on the devices or transmitted with full encryption, so Apple has no access to it. The new AI-powered Siri is initially available only in English on Mac computers in the EU. Apple argues that it has too much market power on iPhones and iPads under the Digital Markets Act (DMA) criteria to demand the same level of data access for competitors as it does for its own systems, and thus decided not to introduce these features on these devices in the European Union for now.

Written by urgent.news from Handelsblatt's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at handelsblatt.com →

More in AI

Parsing JSON from Thinking-Model APIs

When you request structured data from a reasoning model using a JSON mime type, the response often arrives split across several content parts rather than sitting neatly in a single text block.

Chalkboard and avatar for ai agents

Because I find it difficult to read the textual output of agents after each modification or planning, and relate everything to the big picture, I built this simple application (with the help of Claude…

Run DeepSeek V4 on Your Own Hardware With DwarfStar, the Redis Creator's New Inference Engine

The creator of Redis thinks your old GPU is not obsolete. Salvatore Sanfilippo, better known as antirez, published a project called DwarfStar (the repo is antirez/ds4 ) that runs DeepSeek V4 Flash…

  • DeepSeek V4 now runnable on consumer hardware via DwarfStar inference engine.
  • Created by Redis inventor antirez, DwarfStar is a native C program with no GGML dependency.
  • Supports Metal, CUDA, and ROCm backends for optimal performance on various hardware.

More from Sunday 4 October →