As expected, Apple’s WWDC 2026 keynote yesterday morning debuted a new, generative AI-powered Siri that Apple claims can significantly extend the capabilities of any human being with a late model iPhone in their pocket. We’ve heard this before, of course, most infamously in Apple’s WWDC keynote of two years ago. Should we believe that this time around they can actually deliver on the promises they made then? I think not, not at least until they can explain in reasonably granular detail how they plan to address what I still see as three glaringly obvious failure modes for their new agentic Siri.

First, to be even remotely usable in a real-world context, Siri’s promised conversational and multimodal capabilities must be 100% reliable. If they are not, we may wind up being held hostage by a tool that can collect and assesses complex data, draw conclusions from that data, and act on those conclusions too fast for us to predict or escape the consequences of any cognitive mistakes they make along the way.

Second, to do for us what Apple promises, to carry out complex multimodal tasks on on our behalf without any oversight or guidance from us beyond an initial natural language request, the new agentic Siri requires full access to the personal data stored on our computing devices, and full control of the applications running on them. Having presumably overseen the testing necessary to assure that Siri’s responses to our requests can be trusted, Apple seems confident that it can grant Siri this level of access and control to its own applications. For third-party applications, though, the situation is more complex. If, for example, Siri needs to schedule an appointment for an iPhone user, it already knows how to access the data in Apple’s calendar app, and how to add additional data to it. If a user chooses another calendar app as their default, Siri must rely on a feature of Apple’s operating systems called App Intents, which Apple describes as follows:

With App Intents, you express your app’s actions and data in a structured way that makes them discoverable by Apple Intelligence and provides deeper integration with system features people use frequently.

The problem here is that Apple is imposing a new and potentially expensive burden on third-party developers, Small, independent developers may lack the resources necessary to test and deploy Apple’s new guidelines as soon or as thoroughly as Apple expects, and large corporate developers, those who consider themselves Apple’s peers, may prefer their own cross-platform solutions and ignore Apple’s agentic ambitions for Siri as long as they can. (To be fair, in the continuing absence of any but the most rudimentary cross-platform user-interface standards, this seems likely to be the fate of any platform-specific approach to agentic AI deployment, not just Apple’s.)

These frictions between what Apple expects and what will actually happen may eventually reach some sort of accommodation. At this stage in the large-scale deployment of AI agents, though, an ordinary mortal relying on Siri to schedule a ride to a critical medical appointment in another city, transfer complex documents to multiple clients using different computing platforms, or book a two week travel itinerary to cities on another continent, may find themselves in more trouble than anyone, including Apple, is currently able to predict.

Third, it seems clear that very soon our computing devices will not only know everything about us that can be encompassed by digital databases, but will also be able to act in the world as though they were us without any consent on our part other than the implied consent of activating their agentic functions. What happens if we lose control of such a device? Nothing I’ve heard from Apple or any other company has offered anything to convince me that they’ve fully imagined the consequences of millions of digital Doppelgängers suddenly let loose in the world. Are we really sure we want to be continually subjected to an environment in which Apple and Google get to do the fucking around, and we’re left with no choice but to do the finding out? YMMV, but for my part, I’d at least like to be kissed first.