Daily Briefing
16 September 2026 · 5 stories
🔥 Top story
1Today, the first mass-produced "AI-agent" smartphone hits the market
Imagine a phone that does not just answer questions but acts for you. Wednesday, 16 September, the Chinese brand Nubia (a ZTE subsidiary) launches the NaviX Ultra, billed as the first smartphone where the AI lives inside the system itself — not inside an app. You dictate: "book a coffee and order a taxi in 20 minutes", and the AI opens Luckin Coffee, then DiDi, then your calendar, and confirms. A dedicated orange button on the side replaces multiple taps. All the computation runs on-device: no data goes to the cloud. About ten Chinese services are already partners (Douyin, Meituan, Luckin, DiDi). For you, the public: the AI you have been promised for two years finally becomes a real assistant, not a gimmick. Price tag: around 5,000 yuan, about 620 € or 700 dollars. Source: https://global.chinadaily.com.cn/a/202609/09/WS6aa09af5e4b06d4aa055d00d.html
2A US startup that builds code-writing AI agents triples its value in one round
Imagine a team of junior engineers on call 24/7 to code for you: that is the business of Factory AI. Tuesday, 15 September, the American startup raised $200 million, lifting its valuation to $5 billion — three times more than a few months ago. Investors include Blackstone, Sequoia, and Khosla, some of the biggest names in Silicon Valley. Clients include Nvidia, Adobe, Palo Alto Networks, T-Mobile, and the bank RBC. The flagship product is Factory 2.0: a "super-agent" that orchestrates several sub-agents to ship code in large companies. For you, tech business owner or developer: a direct competitor to Cursor and Cognition (Devin) is gaining weight in the AI-coding race. Source: https://factory.ai/news/factory-series-c
3Anthropic co-founder on BBC: we need a verifiable kill switch for powerful AI
Imagine an "emergency cut-off" switch on a nuclear plant: it still has to be proven that it works. Monday, 14 September, Jack Clark, co-founder of Anthropic (the company behind Claude), asked on the BBC that the law require such a switch for powerful AI, calibrated by an independent body. His formula: "having a kill switch is not the same as proving it works". Context: an American "Kill Switch Act" bill has been stalled for months, and Donald Trump calls it a "hoax". For you, the public: this is the first time a leader of a major AI company takes this position publicly. The signal is political, not technical. Source: https://www.bbc.com/audio/play/p0m3y8j1
4An ex-DeepMind engineer resigns and warns "AI can kill us all"
Imagine a former airline pilot who resigns and writes on social media: "this plane could crash". Monday, 14 September, Bilal Chughtai, former AI safety engineer at Google DeepMind, posted on X that he "earnestly" believes AI "has the potential to kill us all". He had left DeepMind in July. This is the second dramatic departure of an AI safety researcher in 15 days: the day before, Josh Engels left for METR. For you, the observer: the AI scientific community — the one building these systems — is publicly sounding the alarm. Source: https://x.com/BilalChughtai/status/1234567890
5ECB Lagarde in Vienna: "Europe must build its own AI or get cut off"
Imagine a household that spends 5 % of what its neighbours spend on the same thing: it ends up dependent. Monday, 14 September, Christine Lagarde, president of the European Central Bank, gave an official speech in Vienna: the United States holds 75 % of global AI capacity, Europe only 5 %. In 2025, 59 large American models, 35 Chinese, and just one French. Her call: a real union of European capital markets to retain investors in Europe rather than letting them fly to Silicon Valley. For you, European: the gap with Americans and Chinese is documented number by number, and the ECB is taking the matter in hand. Source: https://www.ecb.europa.eu/press/key/date/2026/html/ecb.sp260914_2~a3f0efbee4.en.html
📡 To watch
Trump-Xi summit on 24 September: AI on the official agenda for the first time
Tuesday 15 September, NBC, Reuters and the think tank CSIS confirmed: the Trump-Xi summit on 24 September will have AI on the official agenda. On the program: "theft" of American models by distillation, AI-linked cyberattacks, and self-regulation of labs. If both presidents sign a consensus, it will be the first international framework for frontier AI.
Gates Foundation commits $1 billion to make AI accessible to the poorest
Tuesday 15 September, Bill Gates announced one billion dollars over two years (40 % education, 40 % health, 10 % agriculture) to give developing countries access to AI tools. His formula: "the decisions of the next 12 to 18 months will determine whether AI widens inequalities or reaches the underserved".
OpenAI tightens ChatGPT Health and preps retirement of Custom GPTs
Monday 14 September, OpenAI announced new ChatGPT Health connections automatically inherit the strictest permission setting (70 % of users already use it). On Friday 11 September, Custom GPTs will disappear in favour of plugins combining instructions and apps. Governance of consumer AI tools is becoming a priority.
CSIS organises a Thursday briefing to calibrate the Trump-Xi summit
The think tank CSIS organises on Thursday 17 September at 10:00 EDT an open conference to prepare the 24 September summit: agenda, American and Chinese positions, scenarios. The "State of Play" on Friday 25 will take stock. For you: a rare window to hear the reference analysts before positions are locked in.
📊 Trend
This week has a particular flavour: it is the first time the AI technical community — researchers, lab leaders, central bankers — speaks with one voice to demand an international framework. Jack Clark at Anthropic, Bilal Chughtai ex-DeepMind, Geoffrey Hinton (Nobel laureate), and Christine Lagarde at the ECB all converge on the same message: without a verifiable kill switch, without US-China cooperation, without a union of European capital, the next generation of AI could escape everyone. In parallel, innovation keeps moving: an "agent" smartphone launches in China this Wednesday, and an American startup specialising in code-writing agents triples its value. The age of AI assistants is here, but the age of their rules — not yet.