David Bushell’s central complaint is not simply that Apple Intelligence consumes space or appears after an upgrade. It is that a choice he had already made became harder to find after moving to macOS 27. In his account, macOS 15 exposed a control for Apple Intelligence and a separate privacy setting; after the upgrade, the first control was no longer available in the place he expected, while related features were enabled despite his existing “Apple Intelligence & Siri” preference. He says he eventually found restrictions under Screen Time that hide some menus, but that those restrictions do not disable Apple Intelligence itself. [primary]
The article explains that this visible control became harder to find after the macOS 27 upgrade. Credit: dbushell.com.
The article distinguishes this privacy setting from the main Apple Intelligence control and from Screen Time restrictions. Credit: dbushell.com.
The most concrete operational claim is the reported 22.28 GB footprint shown by macOS storage accounting. That figure is evidence about the author’s machine and its configuration, not a universal measurement for every Mac. Storage use can depend on the operating-system build, installed language or model components, account state, and whether features have downloaded data. The source does not provide a controlled comparison with a machine on which the features were never enabled, so readers should treat 22.28 GB as a reported case rather than an Apple-wide estimate. [primary]
The article treats the 22.28 GB figure as a machine-specific observation rather than a universal estimate. Credit: dbushell.com.
The disagreement over the earlier “Report Duration” setting shows why the word “off” needs careful handling. Bushell describes a feature that was “phoning home every 15 minutes with personal data” and says he turned it off. A reader response in the supplied packet challenges that interpretation, saying the setting is a transparency report showing activity associated with Apple Intelligence rather than a switch that independently causes additional transmission to Apple. The packet contains no Apple technical document that resolves that dispute. It therefore supports reporting the author’s observation and interpretation, but not the stronger claim that the setting itself proves a separate 15-minute data transmission. [primary, discussion_cues]
That distinction matters because privacy controls can govern different layers: a feature’s availability, its local processing, its account or device synchronization, its network requests, and the visibility of an activity report. Turning off one layer need not remove every related process, downloaded component, menu entry, or storage allocation. Conversely, a remaining process is not by itself evidence that personal data is being sent. The supplied material does not include packet captures, Apple’s current technical documentation, or a reproducible test matrix showing which actions occur under each setting. Readers looking for a definitive answer should not infer one from the presence or absence of a menu label. [primary, discussion_cues]
The article’s broader relevance is the difference between a technical disablement and a usable refusal. A control buried under parental or Screen Time restrictions may reduce exposure to prompts, but Bushell says it does not disable the underlying capability. That is a meaningful user-experience and governance concern even without assuming malicious intent: if the path to refusal is obscure, users may not know whether they have changed the relevant behavior, and support documentation may not match the interface they see. The evidence here establishes the author’s reported interface experience, not Apple’s design intent or the experience of every user. [primary]
Bushell also connects the interface problem to a wider concern about consent in AI systems, citing headlines about lawsuits and alleged harms involving other companies. Those examples illustrate why readers may care about defaults, data handling, and the ability to withdraw participation. They do not corroborate the specific claims about Apple’s implementation, and the packet does not include the underlying complaints, judgments, technical filings, or company responses needed to assess them. The responsible conclusion is narrower: concerns about AI governance make clear controls important, while the Apple question must be evaluated from Apple’s actual settings and behavior. [primary, support-1, support-2]
For a reader facing the same upgrade, the practical sequence is to separate three questions: what the storage panel says is installed, which Apple Intelligence and Siri controls are visible, and what network or account activity can be independently observed. The supplied evidence can explain why those questions should not be collapsed into a single “off” label. It cannot establish that every hidden process is necessary, that every menu represents an active feature, or that a particular setting has one universal technical meaning across macOS releases. Until those behaviors are documented or tested, the defensible finding is a consent and discoverability problem in the reported experience—not proof of a broader surveillance mechanism. [primary, discussion_cues]