macOS 27 Dropped the AI Toggle, and a Config Profile Puts It Back

on

By the end of this article you will understand how macOS 27's removal of the Apple Intelligence toggle prompted a community-developed workaround, what technical mechanism that workaround uses to operate within System Integrity Protection boundaries, and what the storage and reversibility implications are for Macs in managed or power-user environments.

The toggle is gone

In macOS 27, codenamed Golden Gate, Apple removed the single toggle that previously let users disable Apple Intelligence as a whole. This is a meaningful shift from prior releases, where a user could flip one switch and opt out of the entire feature set. Without that toggle, the on-device AI models remain on disk even when individual AI capabilities are disabled through Settings. The models occupy storage space regardless of whether any feature that consumes them is active. A user who disables every visible AI option in the settings UI still carries the model files on their volume.

A developer going by the handle Om Lahore on GitHub released a tool called RemoveMacAI in response. The tool is open source and, as of the time of writing, has accumulated 1,700 stars on GitHub. Some users on Reddit have confirmed it works. The developer described the tool in a Reddit post as something that turns off Siri, Writing Tools, Genmoji, Image Playground, the ChatGPT extension, and all the summaries, then deletes the models and stops macOS from downloading them again.

Configuration profile, not file surgery

The technically interesting part of RemoveMacAI is how it achieves its effect without violating System Integrity Protection. The tool does not touch anything under /System directly. Instead, it installs a configuration profile that the user approves in System Settings, and it leverages Apple's own asset service to prevent macOS from re-downloading the model files. SIP remains enabled throughout the entire process.

This is architecturally significant for anyone who has worked with macOS in an enterprise context. A configuration profile is the same mechanism Apple uses for Mobile Device Management enrollment and enterprise policy enforcement. By working through that channel, the tool stays within the boundaries Apple has already defined for third-party software to influence system behaviour. It does not require disabling SIP, does not require modifying system binaries, and does not require any operation that a standard user cannot initiate and approve through the settings UI. The developer explicitly stated that nothing in /System gets touched directly.

The reversibility model is equally straightforward. Running the tool's revert command removes the configuration profile, and macOS will re-download the models the next time a feature that depends on them is turned back on. There is no residual state to clean up manually. The system returns to its default posture as if the profile had never been installed. This is the same declarative model that MDM administrators use: apply a profile to enforce a state, remove the profile to revert.

The tool also supports selective removal. A user can disable only the features they do not want while leaving others active. Dictation is unaffected because it operates under a separate setting and is not part of the Apple Intelligence feature bundle that the tool targets.

Storage and operational implications

The developer states that Apple's Intelligence models take up approximately 12 GB of disk space. The Verge has noted that in practice the footprint can exceed 30 GB. For a MacBook Air with a 256 GB or 512 GB drive, that is a non-trivial fraction of total capacity sitting idle if the user never engages with AI features. The fact that the models persist even after individual features are disabled through the settings UI means that the storage cost is not optional in macOS 27 unless an external mechanism intervenes.

For administrators managing fleets of Macs, the absence of a user-facing toggle changes the operational model. Previously, a user could self-serve by flipping a switch. Now, if an organisation wants to ensure Apple Intelligence models are not present on managed devices, the options are more constrained at the user level. The existence of a community tool that operates through a configuration profile suggests that the MDM channel remains a viable path for enterprises to enforce this state at scale, though the source does not confirm whether RemoveMacAI itself is MDM-compatible or whether Apple's own MDM tooling exposes an equivalent payload key.

The fact that RemoveMacAI is a command-line tool rather than a GUI application is also worth noting. It implies that the operation can be scripted and incorporated into provisioning or imaging workflows without requiring a human to interact with a settings panel. For an organisation that wants to standardise a "no Apple Intelligence" baseline across a fleet, a CLI tool that installs a configuration profile is a shape that fits naturally into existing automation pipelines. The developer's description of the mechanism — configuration profile plus asset service, no direct /System modification — is consistent with a design that does not require elevated privileges beyond what a standard user can approve in System Settings.

What the pattern tells us

The broader pattern here is one that sysadmins will recognise from other platforms: when a vendor removes a user-facing control, the underlying mechanism that enforced that control typically still exists in the system. The toggle was a UI layer over a deeper policy mechanism. RemoveMacAI does not exploit a vulnerability or bypass a security boundary; it uses the same policy channel that Apple itself would use if it were to ship a management tool for this purpose. The asset service that macOS uses to download and manage the model files is the same service that a configuration profile can instruct to stop fetching those files.

This distinction matters. A tool that achieves its effect by patching a binary, disabling a security feature, or writing to a protected path is categorically different from one that operates through an API the vendor explicitly exposes for policy enforcement. The latter is reversible, auditable, and does not create a security posture that differs from what the OS was designed to support. For a power user who simply wants the storage back without reading a 40-page MDM deployment guide, RemoveMacAI provides that path. For an organisation, it validates that the configuration profile channel is the correct abstraction layer for controlling Apple Intelligence presence on managed hardware, even in the absence of a user-facing toggle.

Source: arstechnica.com

Comments

Post a Comment

Subscribe to: Post Comments (Atom)

Popular Posts

Install Ollama and Open WebUI on Ubuntu 24.04 with NVIDIA GPU Acceleration (Step-by-Step)

Overview Running a local large language model (LLM) is easier than ever thanks to Ollama and Open WebUI. Ollama handles model downloads and inference, while Open WebUI gives you a clean, chat-style interface in your browser. In this tutorial, you'll install both on Ubuntu 24.04 (works on 22.04 too), enable NVIDIA GPU acceleration, and deploy them with Docker Compose. The result is a fast, private AI stack you control. What You'll Need - Ubuntu 24.04 or 22.04 (fresh or existing server/desktop). - An NVIDIA GPU with recent drivers (Turing/RTX or newer recommended). - Root or sudo access. - Open ports 3000 (Open WebUI) and 11434 (Ollama) on your firewall if you access remotely. Step 1: Install NVIDIA Drivers and Verify CUDA First, update your system and install the recommended NVIDIA driver. On Ubuntu Desktop you can use Additional Drivers, but the CLI route is reliable: sudo apt update && sudo apt -y upgrade sudo ubuntu-drivers autoinstall sudo reboot After reboo...

Install a Local AI Chatbot on Ubuntu 24.04 with Ollama and Open WebUI (Step-by-Step)

Running a fast, private AI chatbot on your own computer or server is easier than ever. In this guide, you will install Ollama (a lightweight local LLM runtime) and Open WebUI (a modern web interface) on Ubuntu 24.04 . You will be able to chat with models like Llama 3 or Mistral without sending data to the cloud, and with optional GPU acceleration if you have an NVIDIA card. What you will need : an Ubuntu 22.04/24.04 machine (VM, bare metal, or WSL), at least 8 GB RAM (16 GB recommended for larger models), 15–30 GB free disk space for models, Internet access, and optional NVIDIA GPU drivers for acceleration. Why Ollama + Open WebUI? Ollama manages local large language models (LLMs) with simple commands and sensible defaults. Open WebUI gives you a clean, chat-style interface with features like prompt history, file uploads (for some models), and model switching. Together, they are a simple, reliable stack for a self-hosted AI experience. 1) Update Ubuntu and install basics First, ...

Install Ollama + Open WebUI on Ubuntu 24.04 with NVIDIA GPU Acceleration (Step-by-Step)

Overview This guide shows you how to install Ollama and Open WebUI on Ubuntu 24.04 with NVIDIA GPU acceleration. You will get a fast local AI stack that can run modern large language models (LLMs) like Llama 3 on your own hardware, secured and accessible in a web browser. We will cover GPU driver setup, Docker with the NVIDIA Container Toolkit, Ollama installation, Open WebUI deployment, basic security, and troubleshooting. Prerequisites - A machine running Ubuntu 24.04 (fresh or updated). - An NVIDIA GPU with at least 8 GB VRAM recommended for 8B models (more is better). - Internet access and sudo privileges. Step 1: Install the NVIDIA GPU Driver Install the recommended proprietary driver so CUDA becomes available to both Ollama and containers. sudo apt update && sudo apt install -y ubuntu-drivers-common sudo ubuntu-drivers autoinstall sudo reboot After reboot, verify the GPU is recognized: nvidia-smi You should see a driver version and your GPU model. If not, check...

Trending Now

Homogeneous Nucleation on Mars: A Textbook Process Finally Observed in Nature

on
By the end of this piece, you will understand what the Arsia Mons Elongated Cloud is, why it resisted explanation for several years after its first observation, and what the newly published model reveals about a physical process that has never been directly confirmed on any planet's atmosphere. The cloud that broke the model Every morning during Martian spring and summer, a thin, extremely long cloud forms near the Arsia Mons volcano and gradually elongates until it stretches to more than 1,000 miles before dissipating. The cycle then repeats for several months. The formation was first recorded by a probe in 2018, and from that point onward it became a persistent puzzle for atmospheric scientists working with Mars data. The cloud's geometry, its timing, and its seasonal recurrence all pointed to a real and reproducible physical process, yet no existing model could reproduce it. For a technical audience familiar with simulation work, the situation is analogous to a product...