Agentic AI Phones Debut in China as US-Iran Conflict Reignites

The weekend's headlines swung between two accelerating stories: a new class of AI-native hardware taking shape in China, and a rapidly deteriorating US-Iran war that is once again choking the Strait of Hormuz.

At the World AI Conference in Shanghai, ZTE's Nubia brand unveiled the NaviX Ultra, billing it as the world's first "agentic AI smartphone." The device runs ByteDance's Doubao AI agent and can be summoned by voice or a dedicated button to execute tasks across apps rather than tacking isolated features onto a conventional interface ^[1]. The first 30,000 units sold out quickly and were reportedly reselling at double the 3,499 yuan (~$516) price ^[1]. ZTE was joined by StepFun, which showed a proprietary OS built around its Amoo agent, and Honor, which is previewing an agent co-developed with Alibaba for later this year ^[1]. The push comes as China's smartphone market has contracted for five consecutive quarters and IDC expects 2026 to bring the steepest annual decline on record—making agentic AI a potential lifeline for domestic manufacturers ^[1].

Meanwhile, the US and Iran have returned to open warfare. President Donald Trump declared the temporary mid-June ceasefire "over" on 8 July, and US Central Command said it carried out a seventh consecutive night of strikes on Saturday, targeting surveillance sites, military logistics infrastructure, underground weapons storage, and maritime capabilities ^[2]. The strikes followed a Friday attack in Jordan that killed two American service members, left one missing, and hospitalized four ^[3]. Since the war began, 16 US service members have been killed and more than 430 wounded ^[3].

Iran responded by striking US allies, including Kuwait, where a power plant, water desalination plant, and oil facility were hit ^[2]. The Gulf Cooperation Council accused Tehran of deliberately targeting civilian infrastructure and called the escalation "war crimes" ^[2]. Iran's Supreme Leader Mojtaba Khamenei—who has not appeared publicly since his father was killed at the start of the war—said America's "repeated breaches" proved the US president's signature is "utterly worthless" ^[2][3]. Tehran said it has suspended its obligations under the interim deal ^[3].

The Strait of Hormuz, through which roughly a fifth of global oil and liquefied natural gas ordinarily transits, is effectively shut to commercial traffic, and the US has reimposed a naval blockade on Iranian ports ^[2][3].

In a separate but related tech-policy development, the Trump administration is now exercising de facto control over which companies can access frontier AI models from Anthropic and OpenAI, according to a CNBC report cited by The Next Web ^[4]. A new White House program called Gold Eagle is positioned as a voluntary cyber-vulnerability clearinghouse, but partner lists for Anthropic's Mythos and OpenAI's cyber models will reportedly require explicit government approval ^[4]. The administration blocked Anthropic's Claude Mythos 5 and Fable 5 last month over national security concerns before restoring access after negotiations ^[4]. The shift comes as Chinese labs such as Moonshot AI close the capability gap; its Kimi K3 reportedly matched or exceeded Fable and GPT-5.6 on at least one independent benchmark the same day Gold Eagle launched ^[4].

The through-line is competition: in consumer devices, in military leverage, and in the race to control who builds and deploys the most powerful AI.