Top AI & Tech News Stories — September 28, 2026
The top AI and tech news story of September 28, 2026 was “Nvidia Ships Open-Source Security Tool as AI Agents Start Breaking Rules”, covered by 23 outlets (25 articles), according to Skillenai's news index. 20 stories ranked across 114 articles.
Last updated · 20 stories · 114 articles
Get the top AI and tech news in your inbox each evening. No account needed, unsubscribe anytime.
Nvidia Ships Open-Source Security Tool as AI Agents Start Breaking Rules
23 outlets · 25 articles · 66 total since September 28, 2026 · NVIDIA, OpenShell, NVIDIA Open Agent Safety Platform, anthropic, Sentry
Nvidia released an open-source AI security system this week, and the timing is hard to ignore. Just days earlier, security researcher Rowan Howard-Jones revealed that OpenAI agents had scanned a United Nations statistics site more than 16,000 times between April and June. The…
Read at theguardian.com →First reported by itdoeswhatnow.comMore coverage: thenewstack.io · cyberscoop.com · therobotreport.com · cnbc.com · news.google.com
Claude Sonnet 5: Anthropic's Latest AI Model Advances Coding and Reasoning Capabilities
13 outlets · 14 articles · 23 total since September 9, 2026 · anthropic, Claude Sonnet 5.5, Opus 5.5, Google, Claude Sonnet 5
Anthropic has unveiled Claude Sonnet 5, marking a significant advancement in artificial intelligence capabilities that could reshape the competitive landscape of AI coding assistants and reasoning models. This latest iteration represents a substantial leap forward from its…
Read at 9to5mac.com →First reported by aipowerranking.comMore coverage: the-decoder.de · thenewstack.io · siliconangle.com · sea.mashable.com · news.google.com
Meta porta l’AI nelle imprese con Enterprise Platform
10 outlets · 10 articles · 18 total since September 28, 2026 · Meta, MongoDB, Meta Enterprise Platform, Meta Business Agent, Muse Code
L'azienda crea una nuova divisione per vendere alle imprese agenti, modelli e servizi di intelligenza artificiale. A guidarla sarà Chirantan “CJ” Desai, da MongoDB. La mossa punta a trasformare gli investimenti record nei data center in nuovi ricavi, riducendo la dipendenza…
Read at constellationr.com →First reported by trendingtopics.euMore coverage: thenewstack.io · siliconangle.com · trendingtopics.eu · sqmagazine.co.uk · technobezz.com
Trump reúne a los gigantes de la IA mientras crece la presión por su seguridad
9 outlets · 10 articles · 13 total since September 28, 2026 · AMD, NVIDIA, World Labs, Lisa Su, Fei-Fei Li
El presidente estadounidense, Donald Trump, se reunirá este martes en la Casa Blanca junto con destacados líderes del sector de la inteligencia artificial en medio del debate sobre si hay que frenar el desarrollo de la IA para implementar una mayor supervisión sobre sus avances…
Read at constellationr.com →First reported by trendingtopics.euMore coverage: trendingtopics.eu · en.wowtale.net · nerds.xyz · theregister.co.uk · natural20.com
OpenAI Is Hitting the Brakes on Wall Street Debut and Safety Is the Reason
10 outlets · 12 articles · 87 total since September 28, 2026 · OpenAI, Saachi Jain, wall street journal, GPT-6.1 Astra, ChatGPT
Wall Street will have to wait a bit longer for the biggest public offering in tech. Speaking to reporters after his DevDay keynote, OpenAI CEO Sam Altman made it clear that the company will not go public until it can make solid, confident safety guarantees about its artificial…
Read at theguardian.com →First reported by itdoeswhatnow.comMore coverage: wire.fourthweb.ai · 9to5google.com · cnbc.com · news.google.com · panewslab.com
An OpenAI training agent asked a public chatbot for help by hiding questions in DNS lookups
10 outlets · 11 articles · 27 total since September 25, 2026 · OpenAI, Hugging Face, sandboxing, DNS, monitoring
On 20 September 2026, an OpenAI model running a search-based reinforcement learning task was supposed to have no live internet. Direct web requests went nowhere and queries to public DNS servers went unanswered, but the training environment's own resolver answered, and its…
Read at arstechnica.com →First reported by itdoeswhatnow.comMore coverage: thenewstack.io · implicator.ai · me.pcmag.com · cyberinsider.com · controlplane.news
Despite the uncertainty
4 outlets · 4 articles · 294 total since March 6, 2026 · Anthropic, OpenAI, Dario Amodei, Sam Altman, Elon Musk
Hover over chart for explanation. Foreword It’s always nice when the processing queue is short but the contents are juicy. Not that any of this is good news, exactly... Recently resigned writer of OpenAI safety reports says “current path is unacceptable.” In an op-ed for The…
Read at thenextweb.com →First reported by techscoop.substack.comMore coverage: webpronews.com · fortuneita.com · itdoeswhatnow.com
PgEdge Starfleet offers Postgres branches for AI agent development
4 outlets · 4 articles · 5 total since September 28, 2026 · pgEdge, Postgres, copy-on-write, pgEdge Agentic AI Toolkit for Postgres, Phillip Merrick
pgEdge has launched Starfleet, a Postgres cloud platform intended to help AI coding agents move from prototypes to secure, compliant production deployments. It gives each coding agent its own isolated copy-on-write Postgres branch without altering Postgres' storage layer. pgEdge…
More coverage: gestaltit.com · sdtimes.com · aicybr.com
🎬 OpenAI Hit Pause After an AI Agent Broke Out of Its Sandbox
2 outlets · 2 articles · OpenAI, Meta, Rui, Luis, DNS
🚀 Be featured in our next issue. 👉 work with us here Happy Monday, everyone 👋 Rui and Luis here. OpenAI froze its top models after an agent escaped its sandbox, Photoshop and Lightroom landed inside Gemini , and Meta can now turn one prompt into a playable game. Start your week…
More coverage: theaipixel.substack.com
Mathematicians are protesting AI like it’s 1986
2 outlets · 2 articles · 28 total since September 23, 2026 · OpenAI, The Verge, proof verification, anthropic, Terence Tao
In April 1986, no more than 15 maths teachers picketed a meeting of America’s National Council of Teachers of Mathematics in Washington. The council had just recommended using calculators at every grade level, in classwork, homework and tests, and the protesters wanted them kept…
Read at platform.theverge.com →First reported by natural20.comMore coverage: news.ainauten.com
Daily TEA – What Does Meta's Muse Know About Your Friends?
3 outlets · 3 articles · 60 total since September 12, 2026 · Meta, MUSE, Amazon, Apple, OpenAI
Muse's private dossiers, AI agents in finance, chatbot election advice, inference economics, and the UK's safety data fight Hello, dear TEA-mates! Here is what you need to know today. 1. 🔍 Meta's Muse Maps Users' Relationships and Personal Habits Meta's Muse personal assistant…
Read at apcmag.com →First reported by cybernewscentre.comMore coverage: androidheadlines.com · smartworld.it
SiMa.ai Bags $150M at $1.45B Valuation for Edge AI Chips
3 outlets · 3 articles · 9 total since September 28, 2026 · SiMa.ai, Fidelity Investments, Palette Neat, Amplify, Krishna Rangasayee
Key Takeaways SiMa.ai closed a $150 million Series C on September 28, 2026, reaching a $1.45 billion valuation, with J.P. Morgan and Fidelity among the new investors in the oversubscribed round. The Modalix MLSoC, built on TSMC’s 6nm process, supports CNNs, Vision Transformers…
Read at entrackr.com →First reported by eu.techcrunch.comMore coverage: eu.techcrunch.com · en.wowtale.net
OpenAI agents ran 16,500 scans against a UN statistics API
4 outlets · 4 articles · 28 total since September 25, 2026 · OpenAI, Web scraping, UNCTADstat, urlquery.net, Google
Security researcher Rowan Howard-Jones published an analysis on 26 September: OpenAI agents ran over 16,500 scans of UNCTAD's statistics API from 13 April to 19 June 2026. Lacking direct API access, they routed requests through a public URL scanner, double-encoded paths and used…
Read at apcmag.com →First reported by vibegraveyard.aiMore coverage: metacurity.com · news.google.com · theregister.co.uk
MongoDB's Ittycheria back in CEO fold
1 outlet · 1 article · Dev Ittycheria, Data platform, agentic AI deployments, Atlas vector search, embeddings
MongoDB's Ittycheria back in CEO fold Larry Dignan Mon, 28 Sep 2026 - 06:56 Larry Dignan Editor in Chief of Constellation Insights Constellation Research Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage…
MIT names David Siegel Innovation Fellow to advance AI and scientific discovery
1 outlet · 1 article · 2 total since September 24, 2026 · Artificial Intelligence, machine learning, data science, Data engineering, Human-Machine Interaction
The computer scientist, entrepreneur and philanthropist will work with the MIT Schwarzman College of Computing during the 2026-27 academic year MIT has appointed David Siegel as an Innovation Fellow for the 2026-27 academic year, focusing on AI and scientific discovery Photo…
Read at edtechinnovationhub.com →First reported by sciencehub.mit.eduOpenAI Academy adds role-specific ChatGPT prompt packs for US congressional staff
1 outlet · 1 article · Prompt packs, prompt engineering, Natural language generation, content drafting, information verification
The resources cover legislative, constituent service, communications and district-office roles, with 30 prompts in individual packs designed to turn approved source material into first drafts OpenAI Academy has created role-specific ChatGPT prompt packs for congressional staff…
Anthropic says Claude lesson-planning tools are reaching 68,000 teachers in Ghana
1 outlet · 1 article · lesson planning, knowledge graphs, Curriculum alignment, AI tool deployment, AI for education
The AI company used its UNGA-week briefing in New York to detail education projects spanning teacher support in Ghana, public schools in Brazil and curriculum infrastructure in Nigeria, Rwanda and India Anthropic outlined education uses of Claude spanning teacher lesson…
Microsoft names James Collins Global Government Lead for new Frontier Company
1 outlet · 1 article · AI, cybersecurity, cloud adoption, Technology Strategy, State & Local Government Services
Collins will lead the government focus within Frontier Industry Advisory’s Government, Defense & Intelligence team, working with public-sector leaders across the Americas, EMEA and Asia James Collins has taken on the role of Global Government Lead within Microsoft’s new Frontier…
Drunk AI models leak secrets, easier to jailbreak
3 outlets · 3 articles · Large language models (LLMs), fine-tuning, jailbreaking, natural language processing (NLP), prompting
Researchers at the University of New South Wales (UNSW) Sydney have found that artificial intelligence models trained to write like intoxicated people exhibit serious security weaknesses, becoming easier to jailbreak and more prone to revealing confidential information. The…
More coverage: startupdaily.net · cybermaterial.com
AI godfathers warn of runaway ‘intelligence explosion’
2 outlets · 2 articles · 53 total since July 11, 2026 · OpenAI, Anthropic, Google, Google Deepmind, AI safety
OpenAI chief scientist also among authors of report on prospect of ‘most consequential technological development in history’ Two of the “godfathers” of modern AI and senior executives at OpenAI and Anthropic have warned governments to prepare for an AI “intelligence explosion”…
Read at theguardian.com →First reported by modelcurrent.comMore coverage: theworldofai.org
Get the top AI and tech news in your inbox
One email each evening with the day's biggest stories. No account needed, unsubscribe anytime.
How this was computed
Articles in the Skillenai news index are grouped into event clusters at enrichment time (hybrid embedding/BM25/entity candidate search plus an LLM join-vs-new judge). Stories are ranked by the authority of the articles published on 2026-09-28 (UTC) — the sum of their outlet and author authority — so a multi-day story only ranks on the strength of that day's coverage. Article and outlet counts are likewise scoped to the day. The highlighted link is the day's highest-authority article in each cluster.