Agentic AI Phones Sell Out in China as US Gates Frontier Models and Iran Conflict Escalates
Three stories are dominating the weekend pulse: China is betting that agentic AI can revive its collapsing smartphone market, the Trump administration is tightening federal control over frontier AI model distribution, and the US-Iran conflict has sharply escalated after American
Three stories are dominating the weekend pulse: China is betting that agentic AI can revive its collapsing smartphone market, the Trump administration is tightening federal control over frontier AI model distribution, and the US-Iran conflict has sharply escalated after American deaths in Jordan.
At the World AI Conference in Shanghai, ZTE unveiled the NaviX Ultra under its Nubia brand, calling it the world's first agentic AI smartphone [1]. The device runs ByteDance's Doubao AI agent and is activated by voice or a dedicated button. The initial batch of 30,000 units sold out quickly and doubled in price on resale markets [1]. StepFun and Honor showed competing agent-first devices, with Honor planning to ship an AI agent co-developed with Alibaba later this year [1]. The strategy is to embed an agentic layer into the OS so AI can execute tasks across apps, rather than stacking isolated features on top of existing interfaces [1]. IDC expects more than half of China's smartphone market could be dominated by AI devices this year as manufacturers seek an escape route from a five-quarter shipment decline [1].
Meanwhile, the White House is asserting de facto distribution authority over frontier AI models from Anthropic and OpenAI, according to a CNBC report cited by The Next Web [2]. The administration blocked Anthropic's Claude Mythos 5 and Fable 5 last month over national security concerns, and the new Gold Eagle programme is expected to put the White House in charge of greenlighting which companies can access new models [2]. A White House official said participation is voluntary, but the operational reality looks closer to a gating mechanism [2]. The timing is awkward: Moonshot AI's Kimi K3 launched the same day and reportedly matched or exceeded Fable and GPT-5.6 on at least one benchmark, prompting former White House AI czar David Sacks to warn that over-regulation could cause the US to lose the AI race [2].
On the world stage, US-Iran hostilities have intensified. Two US service members were killed and one remains missing after Iranian ballistic missile and drone attacks in Jordan on Friday, US Central Command said [3]. The US launched retaliatory strikes on Saturday night targeting Iranian coastal surveillance, air defence facilities, and IRGC forces responsible for the Jordan attack [3]. Iran's army responded with drone strikes on two US bases in Kuwait, according to Iranian state media [3]. The death toll for US forces in the conflict has risen to 16 [3]. A preliminary ceasefire deal struck in June collapsed within weeks, with President Trump declaring it "over" on 8 July [3].
Together, the stories show a weekend where AI hardware competition, AI governance, and Middle East conflict are all accelerating at once.