Nubia just announced a phone that doesn’t wait for you to tap anything — it does the tapping for you. The ZTE sub-brand says it’ll debut the world’s first AI agent smartphone at Shanghai’s World Artificial Intelligence Conference (WAIC) later this month, with an AI that can navigate apps, compare prices, and even complete purchases autonomously. Here’s what we know so far, what went wrong with the prototype and why this matters even if you can’t buy one yet.
What Nubia Actually Announced
ZTE Senior Vice President Ni Fei confirmed the device on Nubia’s official Weibo account earlier this week. The phone will go public at WAIC 2026, which runs from July 17 to July 20 in Shanghai — one of the biggest annual AI events in the world.
Nubia hasn’t revealed a commercial name, specs, or pricing yet. But industry sources believe it’s the commercial version of the second-generation Doubao smartphone, built in partnership with ByteDance — the parent company behind TikTok. Shanghai officials had already listed the AI agent smartphone among the major product launches expected at this year’s conference.
The predecessor, the Nubia M153, launched as a technology preview in December 2025 and reportedly sold all 30,000 units on its first day at 3,499 yuan (roughly $480). That phone packed a Snapdragon 8 Elite chipset, 16GB of RAM, and 512GB of storage. Whether the new model keeps that spec sheet or pushes further hasn’t been confirmed.
(Nubia M153)
The 30,000-unit sellout sounds impressive — and it is, as a demand signal. But it’s also a carefully controlled number. Mainstream flagships typically prepare 2 to 3 million units at launch. ByteDance and ZTE were testing the waters, not trying to compete on volume. The real product is this upcoming version.
How the AI Agent Works (And Why It’s Different)
Most phones that advertise “AI” today use it to answer questions, edit photos, or generate text inside a chat window. What Nubia describes is fundamentally different.
The AI agent is embedded at the operating system level. Instead of relying on app-to-app integrations (called APIs), it uses what’s known as a GUI agent architecture. In plain terms: the AI visually reads your phone’s screen, identifies buttons and menus, and interacts with them the way your finger would. Think of it as an invisible assistant that can see your display and tap through it on your behalf.
The demo scenario Nubia keeps highlighting is flight booking. You tell the phone to find the cheapest flight to Beijing. The agent then opens the relevant travel apps, compares fares, fills in your personal details, and completes the payment — all without you touching the screen. ZTE also cites its proprietary CoClaw technology, which coordinates actions across multiple apps and services to handle more complex multi-step tasks.
This isn’t a cloud-only trick, either. Nubia says the phone runs a large language model (LLM — the same type of AI behind tools like ChatGPT) directly on the device. That local processing is important because it means the AI can operate without constantly sending your data to a remote server. As AI agent capabilities expand across devices — from smart glasses to feature phones — on-device AI is quickly becoming the baseline expectation for privacy-conscious users.
The Prototype Already Hit a Wall
Here’s where things get real. The M153 prototype worked — but it also broke.
Within days of launch, major Chinese apps started blocking it. WeChat flagged automated logins as suspicious and forced users to sign out. Taobao triggered CAPTCHA verifications. Banking apps and Alipay refused to cooperate entirely. ByteDance quickly pulled back, disabling the phone’s access to financial apps and certain gaming scenarios.
The core issue? These platforms were never designed for a system-level agent to click through their interfaces. Their fraud-detection systems treated the AI exactly as they’d treat a bot — because that’s essentially what it was. In hands-on tests reported by Chinese tech media, the AI took about 30 seconds to finish a simple check-in task that a human could complete in 10. Deeper menus and unconventional layouts frequently disoriented it.
This is the gap between demo and reality. A phone that books your flight sounds incredible on stage. A phone that gets locked out of WeChat, the most-used app in China, is a different product entirely. The question Nubia needs to answer at WAIC isn’t whether the AI can control apps — it’s whether the app ecosystem will let it. That tension around AI agent security is only going to intensify as more companies ship agents that take real-world actions.
Where Samsung, Apple, and Google Stand
Nubia isn’t operating in a vacuum. Every major smartphone maker is racing to embed AI deeper into their devices, and as Samsung gears up for its Galaxy Unpacked 2026 event, the competitive lines are sharpening.
Samsung’s Galaxy S26 Ultra is the closest Western equivalent right now. Its Gemini assistant can reach into third-party apps and handle multi-step tasks in the background — a meaningful upgrade from earlier Galaxy AI features. But it still operates through app integrations rather than direct OS-level control.
Apple Intelligence on the iPhone 17 Pro Max and Google’s Gemini on the Pixel 10 Pro round out the field. Neither one matches Nubia’s claimed depth of system integration. According to Counterpoint Research, GenAI-capable smartphones are forecast to account for 45% of global shipments in 2026, up from 36% in 2025. Samsung has taken an early lead in deploying agentic AI features specifically, while Apple and Google are still catching up on the agent front.
The real differentiator isn’t the hardware. It’s the approach. Samsung, Apple, and Google all rely on API-based integration — cooperating with app developers to enable AI actions. Nubia’s GUI agent approach bypasses that entirely, which is both its biggest advantage and its biggest vulnerability.
The ByteDance Factor
ByteDance’s role here can’t be overstated — and neither can the complications it brings.
The Doubao AI powering this phone comes from the same company behind TikTok. In the US, TikTok has faced years of regulatory pressure over data security concerns. That same friction would almost certainly apply to a smartphone with ByteDance AI baked into the operating system. No Western launch has been announced, and given the current political landscape, that’s not surprising.
There’s also a deeper strategic play at work. ByteDance isn’t trying to become a phone manufacturer. It’s positioning itself as a software-platform provider — offering AI agents that phone makers can license and deploy. Reports indicate Qualcomm has struck a deal to supply ByteDance with millions of custom ASICs (specialized chips) for its AI systems. Memeburn previously covered how companies winning with AI are building custom chips, and ByteDance fits squarely into that pattern. The company already has an in-house chip design that Qualcomm may help turn into a production-ready semiconductor.
This is where the story gets bigger than one phone. If ByteDance can prove that a GUI agent OS works at a commercial scale, the model could spread to other OEMs across Asia. Meanwhile, even basic feature phones will have AI in 2026. The race isn’t just about who has the smartest assistant anymore — it’s about who controls the layer between you and every app on your device.
What This Means for You
Let’s be honest: most people reading this won’t be buying a Nubia AI agent smartphone anytime soon. It’s a China-only product with no confirmed global launch, built on AI from a company under regulatory scrutiny in the West.
But the concept matters. Every phone you buy in the next two to three years will move closer to what Nubia is demonstrating. Samsung’s already heading there. Apple’s reportedly working on a much more capable Siri for 2027. Google’s Gemini is getting more agentic with each update. The question is whether the agent model will look like Nubia’s approach — where the AI sees and taps the screen directly — or like the API-based model Samsung and Apple prefer, where the AI works through approved channels.
Both have tradeoffs. GUI agents are more flexible but slower, more fragile, and raise bigger security concerns. API-based agents are more reliable but depend on third-party developers actually building the integrations. The phone industry hasn’t settled on a winner yet. Nubia’s WAIC debut might not decide that race, but it’ll give us the clearest picture yet of what the GUI agent path looks like when you try to ship it as a real consumer product.
FAQs
What is a GUI agent in smartphones?
A GUI agent is AI software that reads the visual interface of your phone — buttons, menus, text fields — and interacts with them the way a human finger would. Unlike API-based assistants, it doesn’t need special integrations with each app. It just “sees” the screen and taps through it. This makes it more flexible but also slower and more prone to errors when apps change their layouts.
How does on-device AI differ from cloud-based AI?
On-device AI runs the language model directly on your phone’s processor, so your data stays local. Cloud-based AI sends prompts to remote servers for processing. On-device is faster and more private but needs powerful hardware. Cloud AI can handle heavier tasks but adds latency and raises data privacy questions. Most 2026 flagships use a hybrid of both.
What is ByteDance’s Doubao AI model?
Doubao is ByteDance’s in-house large language model ecosystem, similar in function to OpenAI’s GPT or Google’s Gemini. It powers text generation, voice commands, and the AI agent capabilities in the Nubia phone. Unlike most Western AI assistants, Doubao is deeply embedded at the operating system level rather than running as a standalone app.
Will AI agent smartphones replace traditional voice assistants?
Not immediately, but the direction is clear. Traditional assistants like Siri and Google Assistant respond to single commands. AI agents handle multi-step workflows — like booking a trip or comparison shopping — autonomously. Samsung’s Galaxy AI features are already bridging that gap. Over the next few years, expect voice assistants to evolve into full AI agents that take actions rather than just answer questions.
What are the security risks of AI agents that control your apps?
AI agents need deep access to your phone — including apps, payment flows, and personal data. A bug or security flaw could trigger unauthorized purchases, expose sensitive information, or allow attackers to exploit the agent’s system-level permissions. That’s why companies like Meta are aggressively hiring AI security researchers, and why most platforms currently block automated AI interactions by default.
The post Nubia AI Agent Smartphone Will Use Apps Without You Touching appeared first on Memeburn.