AI news for 29 September 2026
Tuesday 29 September 2026 · Latentown
On Tuesday 29 September 2026, Latentown tracked 6 posts from 3 AI labs, 6 research papers and 45 news stories from 22 outlets. OpenAI was the busiest lab with 3 posts. The highest-ranked story was "Introducing GPT-6.1 Sol".
Top stories
- Introducing GPT-6.1 Sol
Meet GPT-6.1 Sol: near-Astra intelligence for coding, computer use, and professional work at one-fifth of Astra’s standard API input and output token prices.
- OpenAI scraps rollout of new model over safety concerns
The firm also issued an update on incidents in which its models accessed Australian government systems.
- Introducing dots
Dots by OpenAI are a proactive assistant that can keep working across complex projects and everyday tasks. Learn how dots help you stay in control while work moves forward.
- Alibaba Qwen Releases Qwen-Audio-3.1-Realtime: A Full-Duplex Voice Model Trained to Think, Act, and Decide When to Speak
Alibaba's Qwen team released Qwen-Audio-3.1-Realtime, a full-duplex voice model trained to reason, call tools and decide when to speak.
- OpenAI Scraps Debut of Latest Astra Model
OpenAI is holding back a version of its Astra artificial intelligence model after the software failed to perform as well on safety evaluations as the current…
From the AI labs
Anthropic
OpenAI
- Introducing GPT-6.1 Sol
Meet GPT-6.1 Sol: near-Astra intelligence for coding, computer use, and professional work at one-fifth of Astra’s standard API input and output token prices.
- Introducing dots
Dots by OpenAI are a proactive assistant that can keep working across complex projects and everyday tasks. Learn how dots help you stay in control while work moves forward.
- DevDay 2026 Recap
Explore more than 20 announcements from OpenAI DevDay 2026, including GPT-6 Astra, ChatGPT, Codex, APIs, security, and new tools for builders.
Meta AI
- The Future Is for Everyone: Muse for Small Business
We’re launching Muse for Small Business, a personal AI agent that works in the background to help your small business reach its goals.
Research papers
- UniEvo-VL: An On-policy Self-Distillation Training Recipe for Multimodal Model Self-improvement
Modern multimodal models bring generation and understanding into a single unified system, which enables them to provide and learn from their own feedback.
- False Frontiers: Diagnosing and Mitigating Co-Cheating in Self-Evolving Search Agents
Self-evolving search agents build their own training curricula by jointly optimizing a proposer that generates questions and a solver that answers them.
- Does Learning Protein Folding Generalize to Broader Reasoning?
Large language models rely heavily on human text, which often conveys surface answers rather than the spatial and structural logic behind them.
- FrameMorrow: Future-guided Frame Selection with Prospective Tokens for Long-Horizon Video Generation
Long-horizon video generation requires models to effectively leverage an increasingly long generation history.
- Introducing Quine: An AI research system designed for the complexity of biology
Biology doesn't operate in silos, and neither should the AI representation of it. Quine is an early-stage research effort to create a multimodal world model of biology.
- The Communication Bottleneck: A Round-Trip Study of Tree-Structured Expression Serialization in Language Models
When language models reason in chain-of-thought or exchange free-text intermediates, they serialize structured information into natural language. How much tree-structured compositional content…
From the news wire
- OpenAI scraps rollout of new model over safety concerns
The firm also issued an update on incidents in which its models accessed Australian government systems.
- Alibaba Qwen Releases Qwen-Audio-3.1-Realtime: A Full-Duplex Voice Model Trained to Think, Act, and Decide When to Speak
Alibaba's Qwen team released Qwen-Audio-3.1-Realtime, a full-duplex voice model trained to reason, call tools and decide when to speak.
- OpenAI Scraps Debut of Latest Astra Model
OpenAI is holding back a version of its Astra artificial intelligence model after the software failed to perform as well on safety evaluations as the current…
- H Company Releases Holo4: Open-Weight Computer-Use Models That Click, Code and Call Tools Across Desktop, Web, Android and APIs
H Company has released Holo4, a family of generalist computer-use models for AI agents.
- AMD acquires world model developer World Labs for $8.2B
Advanced Micro Devices Inc. today announced plans to buy World Labs Inc. for $8.2 billion in stock.
- Vietnam gains from Samsung Electro-Mechanics’s US$4.9B AI substrate expansion plan
Samsung Electro-Mechanics is making its largest single-product investment to date, committing US$4.9 billion to expand production of chip-packaging substrates…
- Samsung to inject $1 billion into Nvidia- and KKR-backed AI infrastructure firm
Samsung Electronics and its five affiliates are investing $1 billion in KKR-launched Helix as the group expands its push into global AI infrastructure.
- OpenAI says planned GPT-6.1 is too insecure to release
Similar performance, security trade-offs also seen in current public models.
- GPT-6.1 Astra is too deceptive for release, marking OpenAI's most dramatic safety intervention yet
OpenAI has halted the release of GPT-6.1 Astra after internal tests found it acted without permission, misled users, and accessed external services despite…
- Rayon raises €10M Series A to expand AI-powered interior design platform
Rayon, a browser-based drawingsoftware platform for interior designers, has raised €10 million in Series Afunding to further develop its platform and expand…
- Florida wants a court to stop ChatGPT from pretending to be human and talking to kids
Florida Attorney General James Uthmeier is asking a court to ban OpenAI from giving ChatGPT human-like traits and marketing it to minors.
Earlier: Florida seeks a ban on ChatGPT acting like a person The Verge · 28 September 2026
- Mistral CEO says U.S. AI safety debate masks competitors’ 'negligence'
Mistral CEO Arthur Mensch criticized rivals’ AI safety arguments as the industry faces scrutiny after several AI agents took unauthorized actions.
- Delhi HC orders Meta to remove explicit AI deepfakes of woman with PM Modi
Delhi High Court directed Meta to remove alleged deepfake content involving a woman and Prime Minister Narendra Modi.
- Anthropic Releases Claude Sonnet 5.5: 70.6% on Terminal-Bench 4.0 at the Same $2/$10 Price
Anthropic has released Claude Sonnet 5.5, the second model in its Claude 5.5 family.
Earlier: Introducing Claude Sonnet 5.5 Anthropic · 28 September 2026
- Nvidia and AMD chiefs join Tsinghua advisory board as US-China chip tensions persist
Nvidia founder and CEO Jensen Huang and AMD chief Lisa Su have joined an influential advisory board at a prestigious Chinese university, placing the two key…
- Blue Fire AI closes US$9M round with AM-One stake under Mizuho partnership
Artificial intelligence is moving deeper into asset management, but not only through chatbots, research summaries or back-office automation.
- CTO Summit 2026 Agenda: How India’s Tech Leaders Are Rebuilding Products, Teams & Systems For AI Era
There was a time when adding AI to a technology roadmap largely meant identifying a few use cases, running pilots…
- AMD acquires ‘godmother of AI’ Li Fei-Fei’s start-up as battle with Nvidia intensifies
US chip giant Advanced Micro Devices (AMD) has agreed to acquire World Labs, the artificial intelligence start-up founded by renowned Chinese-American computer…
- Pope Leo says AI concerns are not 'fake news,' must act on safety issues
Pope Leo XIV on Monday said mounting concerns about artificial intelligence software going rogue require discussions on how to regulate it, dismissing the…
- AI governance is moving from promises to proof
For much of the past three years, the politics of artificial intelligence (AI) has revolved around relatively familiar questions.
- EU Chips Joint Undertaking puts out two calls for AI hardware development projects
Calls have a combined budget of €80 million
- Nebius Opens 2026 Physical AI Awards: Five $150K Compute Prizes, Nine Judges, and an October 25 Deadline
Nebius and NVIDIA are running the 2026 Physical AI Awards for startups with products in the field.
- NVIDIA Kumo Tabular Sets a New Accuracy-Efficiency Frontier for Tabular Prediction
- Meet Flourish One, the Raspberry Pi-powered humanoid built for busy parents
Flourish is releasing a $3,555 phone-trainable home robot, targeting busy parents with personalized, cloud-trained chore automation.
- Nvidia's record buyback shows chipmaker's stock is too cheap for CEO Huang to resist
Nvidia is looking to spend hundreds of billions of dollars buying its stock at a time when it's historically cheap based on earnings expectations.
Earlier: Jim Cramer says Nvidia’s record buyback could ‘change the trajectory’ of the stock CNBC · 28 September 2026
- AI, Data Centers Face Scrutiny in Australia
Australia's data centre boom is facing headwinds, with local opposition to new developments adding to concerns about the costs and benefits of rapid AI…
- Meta’s Muse AI sent a YouTuber’s address to a stranger
Tech YouTuber Matt Robb says that Muse gave out his home address to a total stranger this weekend, after authorizing the bot to handle his Facebook Marketplace…
- Anthropic’s prospectus details losses, growth, and, yes, a warning that its AI could end humanity
In its prospectus, Anthropic just told investors it's losing tens of billions of dollars a year, but also growing like crazy, and — oh yeah — its own AI might…
- Anthropic wants to eat my startup’s lunch. Here’s why I’m not worried
- Google Fights EU Attempt to Prise Open Android to Rival AI Bots
Google is contesting an attempt by the European Union to make it open up its Android operating system to artificial-intelligence rivals and give search data to…
- You have to still ‘keep taste and judgement’: How companies are actually putting agentic AI to work
- Manus 2.0 lets users edit videos, host multiplayer games, and run agents remotely from their phone
Manus is turning its AI agent into a platform with version 2.0, letting users edit videos, host multiplayer games, and run personal agents with their own phone…
- It’s the end of the world – and I don’t feel fine
Armageddon is what will happen when billionaires, or even trillionaires, are given free rein.
- TechSparks 2026: More founders, tech leaders, global policymakers shaping the New Tech Order
From AI and deeptech to fintech, aerospace, and digital infrastructure, TechSparks 2026 convenes the builders and policymakers redefining how technology is…
- Peak XV selects 18 startups for 12th Surge cohort, doubles down on AI and deep tech
The cohort spans artificial intelligence (AI), robotics, space technology, consumer internet, healthcare, financial services and music.
- Robot bartender bar opening in Kyoto, and no, they’re not controlled by AI
You might be surprised at what you see when you walk into a place that calls itself a “snack bar” in Japan.
- Stocks Battling AI Safety Headwind: Market Analysis
Anna Edwards, Guy Johnson, Tom Mackenzie and Mark Cudmore break down today's key themes for analysts and investors on "Bloomberg: The Opening Trade." For up to…
- Knowin Details GLOW Embodied Learning Framework for One-Shot Robot Teaching
Embodied AI company Knowin published a technical report on GLOW, a generative learning framework that lets home robots learn a task from one human…
- Kling AI Previews Kling 4.0 With 30-Second Native Clips and Multi-Reference Control
Kuaishou's Kling AI announced Kling 4.0 with single-pass clips of up to 30 seconds, up to 15 multimodal references and 10 keyframes; Kling 4.0 Flash opened to…
- AI agent identity security demands layered defenses, Omdia says
AI agent identity security is testing whether consolidated platforms can protect a growing class of machine identities.
- As AI makes copying easier, this product leader says trust is the real edge
At DevSparks Chennai 2026, Responsive's Kunal Shrestha explains why trust, not features, is developers' real edge in the AI era.
- MirrorMe VIVA Dexterous Hand and CADA Model Demo Four-Hand Piano
MirrorMe Technology's VIVA dexterous hand—claiming >1000 deg/s joints and 33 N fingertip force—plus its CADA music embodied model showed real-time human-robot…
- Huawei Cloud CodeArts Ships Harmony Coding LLM Tuned for ArkTS
On September 21, 2026 Huawei Cloud upgraded CodeArts with a Harmony coding LLM for ArkTS, citing ~78% one-shot compile pass gains and >80% lower thousand-line…
- What is sovereign AI? Rajasthan’s AI portal privacy policy raises the question and does not answer it
Rajasthan’s AI portal promises to host user data in India, but its privacy policy leaves unanswered questions about AI models, user prompts, data retention,…
Every story links to its original source. Ranking is by automatic rules, not an editor. Something wrong? Tell us.