This issue collects 39 AI updates from 2026-09-14, led by Obama urges Democrats to have a clear plan for AI safeguards, AGENTPR unveils AI tool aimed at simplifying media and public reactions, Anthropic researcher's public exit sets off caution talks across AI labs as workers urge a slowdown, Google built a security cage for AI agents on Android, and it is still empty.
Former President Barack Obama said Democrats must make artificial intelligence a central agenda and adopt a very clear plan for the technology's economic and safety risks.Speaking at a Democratic fundraising event, he urged the party to build a framework for a public conversation on AI, arguing the technology can be dangerous if it is not managed.
DW.com reports that Anthropic chief executive Dario Amodei has publicly called for AI development to slow down.ABC News, covering the same discussion, describes a chorus of concern over AI that is growing louder.
The Washington Post reports that Trump is resisting calls to slow down artificial intelligence development at a moment when the political tide around AI is turning.The story is less about any single technology release than about the politics of the American AI policy debate.
House Speaker Mike Johnson told CNN's State of the Union that Congress cannot impose a pause on AI development, saying “China will overlap us” while insisting safety is the industry's own responsibility.President Trump struck the same note the same day, saying the U.S.
The Times reported on 13 September that British ministers avoided imposing constraints on AI, citing the protection of Britain's national interest as the rationale.The report puts the government's trade off between safety concerns and industrial competitiveness back in view, making it a signal for where UK rules on advanced AI are heading.
Microsoft is preparing to release its first code of conduct for its own AI models, and the document is set to be opened to outside comment.Chief executive Satya Nadella has argued that AI governance should not be controlled by a handful of entities but should broadly represent ecosystems, countries and academia.
China's state security minister Chen Yixin warned in a government outlet that advanced U.S.models such as Anthropic's Mythos and OpenAI's GPT 5.
AGENTPR has unveiled an AI tool aimed at simplifying how organizations handle media coverage and public reactions, according to a report by The Guardian Nigeria News.The report gives no pricing, availability or technical detail, so the launch currently stands as a product signal rather than a disclosed capability.
Android Police reports that Google has built a real permission system for autonomous AI agents on Android, centred on a system permission called EXECUTE APP FUNCTIONS and the AppFunctionsManager framework, but it is gated almost entirely to first party software such as Gemini.Because few developers have built the shortcuts it depends on, and Gemini's access is limited to a handpicked tester group, the cage currently controls almost nothing.
Nubia on September 14 released a video showing the core AI features of its NaviX Ultra, a device the report describes as the world's first AI agent smartphone, two days before its September 16 launch and first sale.The video highlights natural voice interaction through an industry first fingerprint AI button, Seed's newest full duplex speech model, support for more than ten dialects, and multimodal screen based Q&A.
CosmosMind, working with more than ten universities including Stanford, Berkeley, MIT, Tsinghua and Peking University, has released MetaRSI v1, which it describes as the first architecture to unify Model RSI, Data RSI and Harness RSI.The team also open sourced its RSI Harness, reporting an average 10.9 point gain for a 3B active small model across four benchmarks and 7.
Peking University spin off Yuankong AI has released Boxer, a mid size on device model, together with two agents: Yuankong AI Work for office tasks and Yuankong AI Science for research.A few days earlier its on device model won a preload deal on HP commercial PCs, with the two sides formally signing a strategic partnership.
Digital Agency Network published a free Agentic Workflows Guide documenting how 10 member agencies deploy AI agents in live client work, with one agency collapsing a week long market intelligence process into a single overnight run.The report also found Anthropic's Claude leading reported usage at 38%, while creative production accounted for just 3% of reported value.
ZTE and telecom operator XLSMART have validated a network optimisation AI agent in live wireless operations, targeting Level 4 autonomous networks.The trial, presented at the ZTE Global Summit in Kuala Lumpur and reported on 14 September, recorded a 25 percent automated closed loop closure rate.
IT Europa reports that Kompingo has sealed a distribution agreement covering EMEA for Reco's agent security offering.Contract value, effective dates and the specific product lines involved are not disclosed, so what is confirmed for now is a channel level move.
A September 14 report describes Oracle Health's Clinical AI Agent as a tool that helps nurses alleviate documentation burden and streamline care.It extends clinical AI from physician note taking into nursing workflows, where the recording load is heavier and more fragmented.
A New York Times report describes how the public resignation of Anthropic researcher Jacob Coxon, citing concerns about the technology, set off internal conversations across OpenAI, Meta and Google, with some AI workers urging executives to pause development until safeguards exist.It also notes that Anthropic and OpenAI are both preparing IPOs and worry that coordinating a slowdown could raise antitrust issues.
Microsoft CEO Satya Nadella announced that xAI's Grok models are being added to Microsoft Copilot, saying the move offers users more model choice, and Elon Musk responded on X to welcome the integration.The rollout initially focuses on Copilot experiences in Word, Excel and PowerPoint, starting with customers in Microsoft's Frontier program.
Zhipu has slipped the first details of its next flagship, GLM 6.0, into a financial filing rather than a technical blog or paper, according to QbitAI.The filing points to a fully self trained approach, while the model itself and its accompanying paper have yet to be released.
DeepSeek now routes all API traffic aimed at V4 Pro to V4.1 Flash and bills it at Flash rates, according to Pandaily.Requests that need V4 Pro capability are therefore answered by V4.
National Business Daily, using the latest OpenRouter data, calculates that global large model token consumption reached 127 trillion tokens in the week of Sept.7 to Sept.13, with Chinese models at 61.
Digiday reports that OpenAI has introduced a new AI native ad format to select clients, attaching a branded business agent to the ad so that a click opens a chat inside ChatGPT instead of sending users to the advertiser's site.Wayfair is trialing the format, and OpenAI CFO Sarah Friar has described today's response ads as only a "basic starting point."
Anthropic's threat intelligence report says its Claude models were used in weapons development, surveillance and cyber operations by actors linked to several countries.The report also flagged five cases involving research that could support biological weapons and a Mali system built to monitor data tied to about 25 million SIM cards.
Chinese consumer electronics brand Ugreen has partnered with Alibaba's Qwen on AI NAS devices aimed at the home, according to a Yicai Global report published on 14 September.The tie up points to a broader push by Chinese model makers to put AI into consumer hardware rather than relying only on cloud APIs.
Ant Lingbo has launched its first embodied large model challenge, built around its LingBot VLA model and open to outside developers and university researchers.The company says the contest is meant to push LingBot VLA deeper into the developer and academic communities and widen participation in embodied AI.
Shengshu has introduced a new world model and paired it with RSI, the idea that a system keeps improving itself without a human setting every next step.The approach aims to solve both where robot training data comes from and who defines the next training goal, while raising fresh questions about verification.
Mashable reports that Nvidia's purchase of Hugging Face has been confirmed at $13 billion.If completed, the deal would place the hub of the open model ecosystem inside the industry's leading AI compute supplier.
Zhipu, a leading Chinese large model developer, has completed a new funding round worth $5 billion, with overseas institutions acting as the main allocators, according to a report from Sina Finance.The investor mix stands out for a domestic model company and has become a reference point for how outside capital views the sector.
Anthropic has told investors it expects positive adjusted operating income for a second consecutive quarter, the Financial Times reported on Sunday, citing people familiar with the matter.The report says gross margins exceed 80% before revenue shared with distribution partners such as Amazon and the cost of training its models, so the figure is not overall profitability.
A Chinese financial AI company has closed a Series B round of more than RMB 300 million, its third fundraising in twelve months, with insurance capital, brokerage backed funds and leading venture investors all taking part.The company says the money will go into model and product development, a broader multi asset client base and overseas expansion.
Airrived announced Agentic Observability on September 14, a capability it says gives enterprises real time visibility and control over every AI agent decision.The Business Wire announcement treats governance of autonomous agents as a product category rather than a compliance afterthought.
Anthropic has taken a dominant position in AI spending among UK small and medium sized businesses, according to FF News.The same figures show SME AI adoption in the country up 1,000% since 2023, suggesting AI purchasing is spreading well beyond large enterprises.
TechCrunch reported on September 14 that A Vinyl Bar in Shibuya, founded by former Spotify innovation head Máuhan M Zonoozy, is releasing experimental music apps as "singles" that put users inside the act of making sound and vocals, backed by a $5.5 million pre seed round.The founder says leaving AI out of each app is an informed choice, because he does not believe the future of music is simply generation.
Superhuman said on September 14 that it is acquiring Y Combinator backed meeting notetaker Fathom, giving its productivity suite a direct feed of meeting topics and action items.Fathom reports more than 400,000 monthly active users and says over 1 million people have recorded meetings on the platform.
A technical breakdown of the DeepSeek V4.1 Flash model card describes a multimodal mixture of experts model with a 552B parameter backbone that activates only about 8B parameters at a time.The weights ship under an MIT license with a 1M token context window, but the model still trails the larger V4 Pro family on several reasoning benchmarks.
PhysBrain 1.5, a physics focused AI model from a Chinese team, has reached the top of a global open source leaderboard, according to QbitAI, which frames the result as clearing the hardest stretch of the physical closed loop.The report puts its spatial intelligence on par with GPT 6 Astra.
China Mobile appeared at the 2026 Computing Power Conference to present what it describes as full stack innovation across computing power and the computing network.The showcase is the carrier's public statement of intent as demand for AI compute keeps climbing.
At the 2026 China International Fair for Trade in Services, China Unicom released an upgraded Xingluo advanced computing power scheduling platform and a new Jingyuan Token service platform in Beijing.Xingluo connects compute pooling with token scheduling through Unicom Cloud 8.0's six compute collaboration architecture, while Jingyuan targets the quality, cost, service and security barriers that keep enterprises from scaling large model use.
The Futurum Group reports that NVIDIA's Groq 3 LPX has entered full production, with the claim of delivering the world's fastest inference as its core pitch.So far the report offers only the production milestone and the performance claim, with deployment customers, availability regions, pricing and benchmark conditions still undisclosed.