Blogger says Apple Intelligence sent data every 15 minutes and macOS 27 removes the opt-out

A blog post on dbushell.com, picked up on the front page of Hacker News on 22 September 2026, documents a two-year stretch of privacy friction with Apple Intelligence on the Mac. It begins with an operating system feature that quietly sent personal data to Apple every 15 minutes and ends, in the author's telling, with a macOS release that no longer offers a complete way to decline it.

What the author found on macOS 15.3

According to the post, the author discovered at 9:51am on Wednesday 5 February 2025 that macOS 15.3 had enabled a feature that contacted Apple's servers every 15 minutes and transmitted personal data. At the time, Apple still shipped a setting that turned Apple Intelligence off. The post notes, however, that a second, related privacy option was tucked deeper into the settings, so fully declining the feature took more digging than the headline toggle suggested.

An upgrade that reversed an explicit choice

The author says they upgraded from macOS 15 to macOS 27 the week before publishing. Despite having "Apple Intelligence & Siri" switched off before the update, the AI features came back enabled, according to the account. More significantly, the author reports that the setting that previously allowed a full opt-out has been removed. Turning off Siri no longer addresses Apple Intelligence, and the post observes that Apple dropped the "Apple Intelligence" wording from the Siri control entirely.

The author also describes hunting through macOS 27 and finding that some AI-related options now sit under Screen Time restrictions, a section intended for parental controls. In their account, those settings do not disable the underlying system; they only tidy up the interface by hiding various AI tools from menus.

Processes that will not die

Even with Siri switched off again, the author says multiple Siri-related processes remain active on macOS 27 and cannot be terminated. The post describes them as consuming memory and writing data, with no explanation of what they are for. That opacity is presented as a problem in itself: users cannot see what the background work is for, and they cannot stop it.

A measurable cost in storage

The post also flags a concrete number: 22.28 GB of disk space attributed to Apple Intelligence after the upgrade. At Apple's storage pricing of £500 per terabyte on new MacBooks, the author works that out to roughly £11 of disk occupied by software they did not agree to install. They call the money trivial next to the consent question, but it is the most easily verified part of the story.

The broader argument

The post's conclusion reaches well past Apple. The author argues that the AI industry as a whole treats consent as an afterthought: capabilities arrive enabled by default, opt-outs are buried, and decisions users have already made are quietly reversed by later upgrades. The closing line is pointed: the author says they were long told to simply turn these features off, and that when they finally did, the option to do so was taken away.

Why it matters

This is one user's account, not an independent audit, and the claims about 15-minute network calls and unkillable processes are not verified elsewhere here. But it records a dated, specific sequence that will feel familiar to many Mac users: a default enabled without being asked, an opt-out that requires effort to find, and a subsequent upgrade that overrides an explicit decision. For admins who manage fleets of Macs, developers who ship on macOS, and anyone drafting privacy policies, it is a useful case study in how operating-system defaults function as policy. It also signals where the industry is heading: as AI is pushed deeper into the platform layer, the boundary between a feature you can decline and infrastructure you cannot keeps shifting, and memory, disk space and network activity are increasingly spent on software the user never requested.