Apple is quietly testing a ChatGPT-like app, though you won’t find it in the App Store. The tech giant has developed an internal tool, codenamed Veritas, as part of the effort to revamp Siri entirely, rather than developing a new chatbot product for consumers, according to Bloomberg’s Mark Gurman.
The app appears to mix Apple’s own large language models with a third-party model, according to people familiar with the matter, who said it is likely integrated with Google Gemini as the company’s partner. The result (which you can see here) mimics the give-and-take style of conversation that ChatGPT popularized — only, this app’s main job is to pressure-test Siri’s new smarts behind the scenes.
What Veritas reportedly does to help rebuild Siri
Veritas acts as a playground for two tentpoles of the Siri overhaul: using personal context and performing in-app actions. That means, in practice, that the assistant needs to be smart enough to know who you’re referring to when you say “send that to Maya,” which of those “budget spreadsheets” you even mean, or how to complete tasks like touching up a photo or pulling details from your calendar without people having to tap their screens.
It is this type of agentic behavior that will require tying together multiple systems — language models, on-device signals, app intents, and strict privacy controls. Apple has already hinted at elements of this vision under its Apple Intelligence umbrella, and detailed a privacy approach it calls Private Cloud Compute for any requests too heavy to be processed on-device. Veritas gives Apple employees a controlled environment to test accuracy, latency, and safety before those features appear in Siri.
Apple, however, has historically dogfooded major platform shifts internally. Past acquisitions and experiments, such as Workflow evolving to Shortcuts, are prime examples of Apple using internal power users to incubate powerful capabilities before integrating them deeply at the system level.
However, Apple might keep it internal for various reasons. It is about quality and safety. General-purpose chatbots can delude, and companies remain exposed to brand and legal dangers. Apple’s bar for reliability is very high, particularly when such features touch personal data.
An internal app can be altered aggressively without creating the expectation that it is supposed to behave like a finished product.
On the other hand, this involves economics and scale. In fact, analysts have reported that running large-model inference in the cloud can cost a few cents per query, multiplied across an installed base of more than two billion active devices.
Apple’s approach emphasizes on-device modeling for speed and privacy, yet a hybrid is necessary for complex requests.
Besides, it is about coherence. Apple’s substantial value proposition is not “yet another chatbot” but a system-level assistant able to understand the devices, apps, and context. Packing a distinct chatbot in might have muddied the roadmap.
Turning the instruction from Veritas into a dedicated Siri experience keeps the story clean. So, it indicates a direction for Siri and Apple Intelligence. Apple has recognized that rebuilding Siri’s base will strengthen it. The company’s software chief has outlined a new architecture designed to align natural-language understanding with app actions and data promises — precisely the area Veritas is expected to review.
There are also indications that Apple is delving into some closely related AI services. Bloomberg says that a “web-based version of the answer machine,” to act as a “yellow pages” for Siri, is currently in development. Add that to selective use of partner models and the Apple Intelligence layer, and you can see the trend: Siri as an orchestrator that dispatches requests to the best tool for the job rather than a monolithic voice bot trying to do everything.
The competitive context for Apple’s evolving Siri plans
Rivals are moving fast. Google is integrating Gemini with Assistant and Search, Microsoft is making Copilot a feature across Windows and Office, and Samsung trumpets device-level “Galaxy AI” capabilities. The frontier is no longer chat per se; it’s context-aware agents that carry out tasks on your behalf within apps and across services.
Apple’s differentiator is tight hardware-software integration and a privacy stance that tries to keep as much computation on your device as it can. If Veritas helps Apple ship a Siri that is capable of, say, finding the particular PDF your accountant sent and dropping its totals into a budgeting app with one request — reliably and privately — it will matter more than debuting yet another me-too chatbot.
The bottom line for iPhone users: expect smarter Siri
Don’t expect to download Veritas. Do expect its lessons to manifest in a smarter Siri with much richer, cross-app actions and enhanced insight into your context. The app is a laboratory, not a product — evidence that Apple is now prioritizing substance over sizzle as it rushes to make its assistant actually useful once more.
