On Tuesday, Apple released the public beta of iOS 27, putting its revamped AI-powered Siri in the hands of anyone with a compatible iPhone. No developer profile is needed: the over-the-air update delivers an assistant designed to better understand natural language, handle multi-step requests, and integrate deeply with apps.

It was a forced move for Apple. After years of Siri lagging behind Alexa and especially the ChatGPT ecosystem, the company needed to mark its territory. The public beta isn’t just a sneak peek for impatient users—it signals that AI is now an operating-system-level commodity.

Who wins and who loses? End users gain a smoother, more capable assistant. Developers get a broader testing ground for integrations. Google and Samsung, which have invested heavily in generative AI on smartphones, face stiffer competition. But the overlooked point concerns organizations that think in terms of data sovereignty. Siri AI, even when it runs parts of the processing on Apple’s local Neural Engine hardware, structurally depends on cloud services for complex inference and up-to-date knowledge. It’s a classic hybrid AI: on-device for basic privacy, cloud for power. That compromise might suit consumers, but it clashes with enterprises that must keep data under their physical control, without a single interaction reaching third-party servers.

The public beta also acts as a litmus test for hardware evolution. Apple touts proprietary chips capable of low-power neural execution, but provides no figures on VRAM load, numerical precision, or context window. The silence is deliberate: in Apple’s model, these metrics are opaque by design, whereas anyone building on-premise or self-hosted LLMs needs full transparency around quantization, throughput, and resource consumption. The difference is almost philosophical—on one side, a polished but boxed-in experience, on the other, granular control demanding open frameworks and commodity hardware.

A second-order effect revolves around trust. The more Apple pushes AI as an iPhone differentiator, the more awareness grows that the average user has no idea where their data ends up. Yet that’s precisely the battlefield for enterprise adoption. If a voice assistant can’t enter an operating room, a law office, or an air-gapped factory because it constantly phones home to the cloud, then the path to truly business-integrated AI is paved with different solutions. It’s no coincidence that self-hosted alternatives are gaining ground among companies that evaluate total TCO, even if they require hardware investments and expertise that Apple, by its very nature, will never provide.