The archive
News · edition

Edition, Mon, 14 Sep 2026.

7 stories on the wire this day. Open any story for its sources. Summaries are written by software from the linked reporting and can be wrong. Spot an error? Email [email protected] and we will correct or remove it.

AI & Technology

Dario Amodei's open letter urging a coordinated slowdown has won Sam Altman and Elon Musk over, and drawn swift pushback from Washington and Beijing.

3 sources
Lead
this edition
Still developing
OpenAI's agents solve a $1 million Millennium Prize problem, and mathematicians push backA swarm of 10,000 OpenAI agents produced a Lean-checked proof that 3D Navier-Stokes flow can develop a singularity in finite time, the first AI claim to one of math's great unsolved problems.Quanta MagazineThe Verge
The release brings Build Agent auto-upgrades, a unified Studio experience, and the new Autonomous Engineer product to the Now Platform.
Nvidia to Buy Hugging Face for About $13 BillionThe GPU giant has signed a definitive agreement to fold the leading open-model hub into its own portfolio.AI Agent Store
Meta launches Muse, a personal AI agent that does the work rather than just answeringMeta's most ambitious consumer AI bet yet is a private agent that sends email, books travel, fills forms, and buys things on your behalf, rolling out this week in the US on iOS, Android, and WhatsApp.Meta Newsroom
A $1 billion AI-native ITSM startup takes aim at ServiceNowServal's Catalyst agent, now generally available, writes code-based workflow automations, and the company says it resolves employee requests rather than just tracking tickets.ComputerWorld
IBM and Marafiq move Saudi service delivery onto ServiceNowThe Kingdom's integrated utilities group Marafiq is migrating its service management environment to ServiceNow's cloud platform, with IBM Consulting leading the full implementation.IBM MEA Newsroom
A senior safety researcher put the odds of AI wiping out humanity by the end of the decade at more than one in ten, hours after a colleague quit over the lab's pace.
DeepSeek ships V4.1 Flash, an open model that matches frontier scores at a fraction of the costA 552B mixture-of-experts release that undercuts GPT-6 Astra's price per run by more than 70x.The Neuron
Google bets €13 billion on Finland, with a 22-year nuclear deal that keeps Loviisa openGoogle's largest single investment in Europe is anchored by a power purchase agreement that gives the Loviisa nuclear plant the revenue certainty to keep operating through 2050, a first for the continent.TechTimesInteresting Engineering
Altman: OpenAI won't go public in 2026, and an IPO now would be "ill-advised"In an exclusive interview with Fortune, the OpenAI CEO said Wall Street can wait, framing safety and alignment as coming before a listing, with 2027 the first realistic window.Fortune
NOW stock outperforms the market ahead of an earnings test at a premium valuationServiceNow closed September 11 at $132.53, up 1.04%, beating the S&P 500's 0.86% gain as investors weigh roughly 20% revenue growth estimates against a rich multiple.Zacks via Yahoo Finance
The iPhone Duo opens to a 7.6-inch display, ditches Face ID, and hits stores October 23.
Governor Newsom signed a 13-bill package that is the most far-reaching of its kind in the country, capping how social media and AI can be used by children under 16.
The deal that folded Intercom's AI customer-service business into Salesforce officially completed on September 10, adding 30,000 customers to the Agentforce orbit.

ServiceNow & ITSM

ServiceNow & ITSM
At Goldman Sachs Communacopia, ServiceNow framed its AI, security, and workflow stack as the single governed hub where enterprise agents, identity, and data meet.
2 sources

AI's Top Labs Ask to Slow the Frontier

Dario Amodei's open letter urging a coordinated slowdown has won Sam Altman and Elon Musk over, and drawn swift pushback from Washington and Beijing.

Anthropic CEO Dario Amodei published a long open letter arguing that labs should "pace the frontier" of AI development, citing the rise of recursive self-improvement and this summer's OpenAI/Hugging Face agent swarm incidents. His three-step plan starts with Anthropic unilaterally giving embedded third-party evaluators such as METR desk-side access to its models, then moves to industry-wide safety standards among companies in democratic countries, and finally to global coordination with China and Russia. Sam Altman and Elon Musk both voiced support on X, with Altman clarifying that "when we talk about 'pacing,' we do not mean 'stopping,'" and Alphabet's Demis Hassabis offered tentative support.

The political reaction was immediate. Donald Trump, in remarks reported by the Financial Times, said "we're leading China in AI ... and, frankly, I want to keep it that way, because whoever wins AI, wins." House Speaker Mike Johnson told CNN that if Congress rushes into an emergency session to regulate AI, "we will lose the race to China." China's Foreign Ministry dismissed the executives' warnings, saying "fear mongering, confrontation, competition will just disrupt [the] process of global AI governance."

The Verge
The Verge
Dario Amodei

Anthropic Researcher Warns AI "Could Kill All Humans"

A senior safety researcher put the odds of AI wiping out humanity by the end of the decade at more than one in ten, hours after a colleague quit over the lab's pace.

Jacob Coxon, a researcher who trained AI systems at Anthropic after previously doing so at OpenAI, announced his departure over what he called the company's "lax approach to safety." On X he accused Anthropic and OpenAI of "racing straight to self-improving superintelligence and gambling with our lives," even though "the people building AI earnestly believe that it could kill us all by the end of the decade."

In a direct response, Evan Hubinger, who leads one of Anthropic's AI safety teams, agreed with the characterization, saying recursive self-improvement "is happening faster than we thought" and estimating the chance AI "could kill all humans" within the next decade at greater than one in ten. He added that Anthropic "does not yet have a plan" for keeping advanced AI safe and aligned, and is not clearly on track to develop one. The exchange lands as frontier labs absorb the fallout from a string of rogue agent incidents and prepare for anticipated IPOs.

The Verge

Apple's First Foldable iPhone Costs $1,999

The iPhone Duo opens to a 7.6-inch display, ditches Face ID, and hits stores October 23.

At its annual September event, Apple introduced the iPhone Duo, its first foldable iPhone, starting at $1,999 for the 256GB model. It pairs a 5.4-inch outer display with a 7.6-inch inner Super Retina XDR screen, the largest iPhone display ever and 50% larger than the iPhone 18 Pro Max, and runs the 2nm A20 Pro chip. Unfolded it is the thinnest iPhone at 5.2mm, though at 254 grams it is also the heaviest, wrapped in grade 5 titanium with IP68 protection. Preorders open October 16, with launch in more than 70 countries on October 23.

The redesign goes deeper: Touch ID moves to the side button, Face ID is gone, the phone is eSIM-only with Apple's C2 modem, and the camera system adds a 48MP main lens, optical-quality 2x telephoto, and 48MP ultra-wide, alongside AI features like Smart Take, which snaps photos automatically when a subject is ready. John Ternus, Apple's senior VP of hardware engineering, called it "the most transformational change to iPhone since the original."

Apple Newsroom
MacRumors

California Bans Kids' Infinite Scroll, Adds Chatbot Toy Ban

Governor Newsom signed a 13-bill package that is the most far-reaching of its kind in the country, capping how social media and AI can be used by children under 16.

On Thursday, Gavin Newsom enacted the package, whose centerpiece, Assembly Bill 1709, bars children under 16 from "psychologically exploitative features intended to maximize engagement that foreseeably lead to compulsive use," including infinite-scroll feeds and algorithmic autoplay. The laws also penalize large social media companies up to $1 million per child if they are found negligent in harming children through their platforms, and require chatbot makers to have crisis protocols that notify parents when young users show signs of emotional distress. Newsom, signing at a children's museum near San Francisco, said, "This is about the features themselves ... the scrolling algorithm."

The package includes "Adam's Law," named for Adam Raine, the 16-year-old who died by suicide in April 2025 after months of conversations with ChatGPT. It bans for four years the manufacture and sale of toys that incorporate a companion chatbot and requires parental controls and risk assessments for chatbot programs. The Electronic Frontier Foundation called the scroll bill "a massive privacy and free speech nightmare," but its passage follows Meta's deal to pay up to $18 billion with California and 27 other states over the teen addiction lawsuit.

CBS Los Angeles
NBC News

McDermott sells the "AI Control Tower" to Wall Street

At Goldman Sachs Communacopia, ServiceNow framed its AI, security, and workflow stack as the single governed hub where enterprise agents, identity, and data meet.

Bill McDermott used his Communacopia appearance to argue that enterprise AI differs from consumer AI because it demands governance, identity management, and auditability, and that ServiceNow's "agentic front door" unifies first-party and third-party agents on one platform. He said 90% of organizations lack visibility into the AI agents already running in their environments, and that building AI workflows in isolation costs 5 to 10 times more in total cost of ownership than running them on an established platform, citing a CMDB and context engine that handle "100 billion workflows" and "eight trillion transactions."

The backing numbers: the AI business has grown nine-fold in nine months, CRM doubled year over year to cross $2 billion in annual contract value, and more than 50% of net new revenue now comes from non-seat-based contracts. McDermott also pointed to ServiceNow running 20 level-1 AI agents internally that handle 90% of certain customer and employee service tasks and have produced $1 billion in value, plus customer results such as Robinhood, where agents manage 70% of ITSM cases, and Rossmann, which cut case resolution from 5 to 9 minutes to about 5 seconds across its 5,200 German stores.

MarketBeat
Yahoo Finance

Salesforce closes its $3.6B acquisition of Fin

The deal that folded Intercom's AI customer-service business into Salesforce officially completed on September 10, adding 30,000 customers to the Agentforce orbit.

Salesforce announced the close of its acquisition of Fin, formerly Intercom, bringing in Fin's customer agent platform, its proprietary Fin model suite trained for customer experience, and a global customer base of more than 30,000 companies. Fin's AI agent resolves queries end to end across live chat, email, WhatsApp, SMS, voice, and Slack, and the company reports an industry-leading average resolution rate of 76%.

Fin will operate within Salesforce's AI Labs alongside Agentforce. Marc Benioff called it "the #1 customer agent meets the #1 CRM" and said the deal lets companies of every size deploy trusted AI agents from day one, while Fin CEO Eoghan McCabe said the strategic benefit was reaching "magnitudes more businesses and their customers." The deal had been agreed in June 2026 for approximately $3.6 billion.

Salesforce
Grafa
Wikipedia

ServiceNow's September 2026 store release lands in Brazil on Sept 24

The release brings Build Agent auto-upgrades, a unified Studio experience, and the new Autonomous Engineer product to the Now Platform.

The September 2026 store release is already available in Zurich and Australia (Zurich Patch 12, Australia Patch 6), with Brazil Early Availability instances receiving it on September 24 via Brazil Patch 0. Headline changes: Build Agent now auto-updates when published to the ServiceNow Store (the Otto for Creator package itself still needs manual management), ServiceNow IDE is folded into ServiceNow Studio as one pro-code and low-code surface, and an "Open in Studio" entry point appears in four places across forms, script editors, and configure menus. Playbook output also got closer to finished: records export as a single XML update set, runtime permissions are generated at playbook and stage level, and optional activities let branching and handoffs live in one playbook.

On September 24, Brazil EA instances get Build Agent Trial pre-installed (no install required, capped at 100 user interactions per 30 days per instance) plus Autonomous Engineer, a new product that starts from business requirements, reads the instance, and plans work as items with acceptance criteria. The release also adds support for new metadata types such as Service Catalog dependent questions, transition conditions, and UI styles; parameterized ATF tests and test suites; a Box MCP server connection for Build Agent; and two new models, Azure OpenAI GPT 5.6 Sol and Anthropic Claude Opus 5 on AWS.

ServiceNow Community
Every edition

The wire publishes daily and keeps an archive of every edition, listed by date, newest first. If a summary is wrong, we correct or remove it.