
Apple has been teasing “Apple Intelligence” all year, but most of what users have so far are basic AI features. Think text rewriting, email summaries, Genmoji, live translation and those so-so generative images. Impressive for Apple’s ecosystem, sure, but still not anywhere near the assistant experience that users see when asking ChatGPT, Gemini or even Perplexity for help.
Now, Apple is finally showing signs that it wants to fix Siri once and for all. And the solution involves a secret internal chatbot.
According to Bloomberg, Apple has quietly built a ChatGPT-like iPhone app called Veritas (Latin for “truth”) to test the new Siri that’s scheduled to launch as early as March 2026. The app isn’t meant for public use. Instead, Apple employees across its AI division are using it to experiment with new Siri features, get feedback faster and iron out failures before shipping anything to customers.
Veritas reportedly looks and feels like most chatbots: employees can start different conversations, follow up on past queries, and ask complex questions. Apple is using it to test Siri’s ability to search personal info like songs and emails, carry out in-app actions such as editing photos, and reference previous chats — things the current Siri simply can’t do without punting the user to ChatGPT or a web search.
A make-or-break comeback attempt
The revamped Siri, powered by a new system code-named Linwood, relies on large language models (LLMs) from Apple’s own Foundation Models team, plus a third-party model. And here’s where things get interesting.
It’s been reported before that Apple has been negotiating with outside AI players like OpenAI and Anthropic. But recently, Apple has turned its attention to Google, exploring a custom version of Gemini to run parts of Siri’s intelligence under the hood. That possibility makes Veritas more than just a testbed. It’s Apple figuring out whether its own tech is enough, or if it needs a little help from the neighbours in Mountain View.
Apple had originally planned to ship its Siri overhaul earlier this year but delayed after features failed almost a third of the time. The internal chaos even led to a shakeup at the top of its AI group. Now, with CEO Tim Cook reportedly telling staff that AI is “ours to grab,” pressure is high to get it right. If the upgraded Siri lands smoothly in March, Apple could finally show that it has a real AI strategy beyond Genmoji and photo edits. If not, the company risks sliding even further behind Samsung and Google, both of which are integrating AI deeper into everyday smartphone tasks.
Long overdue upgrade
For users, what matters most is this: Siri still can’t understand a wide range of questions without becoming confused or passing the request to ChatGPT. It can’t intelligently extract details from your apps or screen, and it doesn’t handle follow-up chats the way modern AI assistants do. Those limitations are glaring in 2025.
That’s why Veritas is significant. Not because Apple will release it, but because it signals that the company is finally prototyping a Siri that behaves like ChatGPT or Gemini, but lives right inside the iPhone.
And it won’t stop there. Apple is also working on a visual redesign for Siri next year, new AI-powered smart home devices, and expanded AI features for the HomePod and Apple TV. It’s even making a bigger push into AI-based web search, another area where Google looms large.
Too late or right on time?
AI features are quickly becoming a deciding factor when buying a phone. Yet during the iPhone 17 launch, Apple barely mentioned its own AI ambitions. Compare that to how Samsung pushes Galaxy AI or how Google treats Gemini as a core Android experience.
With Veritas, Apple is testing whether a chatbot-style approach is the fastest way to fix Siri and rebuild trust in its AI direction. The company doesn’t plan to release the app to the public, but if the technology inside it works, you’ll eventually see it baked right into the new Siri.
And if part of that experience runs on Google Gemini? Well… Apple users might not care, as long as Siri finally stops pretending it doesn’t understand the question.
Discover more from Techish Kenya
Subscribe to get the latest posts sent to your email.