Singapore AI News & Daily Briefing

Trending: ###### [Exclusive! 11th Annual Malaysia OpenGov CXO Leadership Forum 2026 – Fireside Chat with YB Gobind Singh Deo, Minister of Digital](https://opengovasia.com/exclusive-11th-annual-malaysia-opengov-cxo-leadership-forum-2026-fireside-chat-with-yb-gobind-singh-deo-minister-of-digital/) ###### [Exclusive! 11th Annual Malaysia OpenGov CXO Leadership Forum 2026 – Technology in Motion](https://opengovasia.com/exclusive-11th-annual-malaysia-opengov-cxo-leadership-forum-2026-technology-in-motion/) ###### [Exclusive! 11th Annual Malaysia OpenGov CXO Leadership Forum 2026](https://opengovasia.com/exclusive-11th-annual-malaysia-opengov-cxo-leadership-forum-2026/) ###### [From Research to Reality: Singapore’s Quantum-Safe Testbed Recognised at Digital@UNGA 2026](https://opengovasia.com/fro

**SINGAPORE, 5 October 2026** – [Singapore’s Digital Economy grew to S$144.1 billion in 2025](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/sgde-ar-2026), accounting for 19.3% of the nation’s gross domestic product (GDP), up from 18.8% in 2024. This steady growth reflects that digital technologies continue to shape how all enterprises across various industries operate, and how workers do their jobs. * According to [IMDA’s Singapore Digital Economy (SGDE) Report 2026](https://www.imda.gov.sg/assets/8e4d9951-bc53-4612-80ba-77ee4feaa6da.pdf), non-Information & Communications (non-I&C) sectors remain the dominant growth driver, contributing to more than two-thirds of Singapore’s digital economy. The nominal value-added from digitalisation in the non-I&C sectors grew

A crucial safeguard against AI agents going rogue —keeping humans in the loop to review and approve their decisions—will fail unless designers and users change their current practices, a trio of leading AI ethics researchers argue. Though most autonomous agents have systems to keep users in the loop about their actions, in practice these processes actually push humans out of the loop, the authors argue in a paper posted to ArXiv on 6 September. In other words, “the human just becomes this meat tool to give permissions without the cognitive capability to engage,” says one of the authors, Avijit Ghosh , the lead technical AI policy researcher at Hugging Face, an open-source machine learning platform. In the near term, the paper says, humans being out of the loop leads to agents acting in way

On 13 September, leading luminaries and up-and-coming talents in mathematics and computer science congregated at the annual Heidelberg Laureate Forum in Germany for a week of discussion, networking, and—naturally—wurst. Having attended several of these events in the past, the mathematicians normally focus their chatter on which famed researchers are present, or what interesting problems they have been working on. But instead, every snippet of conversation caught in passing or any debate accidentally overheard was about how AI companies such as OpenAI, Anthropic, and Google are steamrollering their way through mathematics. And there is good reason for these fervent discussions. Mathematics is the perfect testing ground for AI, involving step-by-step logical reasoning and answers that are au

For the first 10 years of my career, I worked in product management and data analytics by myself. I wrote database queries that pulled numbers out of corporate systems, built statistical models to predict what customers would buy, and shipped data pipelines that moved information between business systems. I built and scaled analytics teams at Best Buy and Target , studying how customers shop and what stores should stock. Today I lead enterprise AI transformation at Lowe’s , the Fortune 100 home improvement retailer. The goal is not to sell artificial intelligence ; it is to use it to deliver useful expertise at the moment a customer needs it. In retail and other customer-facing industries, virtual assistants can help people address everyday questions—such as how to repair a leaky faucet—wh

Space was always supposed to be the final frontier of human exploration. It’s shaping up to be the final frontier for artificial intelligence too. Last December, NASA’s Jet Propulsion Laboratory used Anthropic’s Claude models to help plan two Mars drives for the Perseverance rover , with human planners checking and adjusting the route before upload. In May, NASA and IBM put a compressed AI model on the International Space Station and a satellite to identify things like floods and clouds from orbit, the first model of its kind demonstrated in space. And in July, astronauts on the ISS tested a large language model to see if it could help with questions on maintenance procedures . These experiments point to a larger shift in space engineering. For decades, engineers on Earth determined what a

We’re introducing a new visual ad format in ChatGPT, expanding our measurement tools and partnerships, and developing new ways to understand brand suitability. ChatGPT reaches 1.2 billion people each week, making it the world’s largest AI-native consumer platform. Advertising helps us bring the benefits of AI to more people. That broad access is central to our mission, and a strength of our business. It allows us to serve a much larger market than subscriptions alone, while giving businesses a new way to reach people as they explore ideas, make decisions and discover products and services relevant to what they’re trying to accomplish. ## Bringing products and services to life Our new visual ad format helps people imagine how products and services could fit into their lives. Through images

[Content provenance](https://openai.com/index/advancing-content-provenance/) helps people understand where content came from, how it was created or edited, and whether it contains signals associated with our models. We’ve already made tools publicly available to identify images and audio generated by our models. Today, we’re sharing our approach to text watermarking in response to the EU AI Act, and how it fits into our broader work. The EU AI Act requires generative AI providers to make generated text identifiable in a machine-readable way. Text watermarking and detection remain early technologies with significant limitations, and views about their benefits and responsible uses are still developing. Our phased approach reflects both the EU AI Act requirements as well as the technology’s l

This White Paper gives data engineers and architects a practical overview of how parallel partitioned reads, write-path optimization, and cloud-native bulk loading reduce large-table replication times, and why replication speed has become a business concern as data volumes grow. Download this free whitepaper now!

[**TechCrunch Disrupt 2026**](https://techcrunch.com/events/techcrunch-disrupt/?utm_source=tc&utm_medium=post&utm_campaign=disrupt2026&utm_content=ticketsales&promo=post_3daystildisrupt_10102026&display=) starts in just three days. Moscone West is getting ready, hundreds of startups are preparing to show what they’ve built, and thousands of founders, investors, and tech leaders are heading to San Francisco. The three-day countdown is on, and so are your pass savings of up to $100. [**Grab your pass**](https://techcrunch.com/events/techcrunch-disrupt/?utm_source=tc&utm_medium=post&utm_campaign=disrupt2026&utm_content=ticketsales&promo=post_3daystildisrupt_10102026&display=) before TechCrunch Disrupt 2026 doors open on October 13 at 8 a.m. PT. Bring a co-founder, colleague, or peer and [**ge

Singapore is positioning itself as a regional testbed for trusted voice AI, aiming to set standards for how voice assistants are built and deployed safely. The push matters because voice AI is spreading fast into banking, customer service, and government hotlines, where mistakes or deepfake-style impersonation can hurt real people. If the testbed works, Singapore businesses and agencies could get clearer rules and more reliable voice tools faster than their regional peers.

Chinese EV maker Xpeng has started mass-producing robotaxis powered by its own in-house Turing AI chips, with each car using four of them. The vehicles are designed for Level 4 autonomy — no human driver needed in set areas — with pilot operations planned for late 2026 and safety officers removed by early 2027. Xpeng turned to its own chips after delays with Nvidia's processor, joining other Chinese automakers cutting reliance on US semiconductors amid export-control worries. This puts it in direct competition with Tesla's self-driving push and signals that custom AI chips are becoming central to the robotaxi race.

[TechCrunch Disrupt 2026](https://techcrunch.com/events/techcrunch-disrupt/?utm_source=tc&utm_medium=post&utm_campaign=disrupt2026&utm_content=ticketsales&promo=foundersguide) is built around one question: How do you build an enduring company in the AI era? Our programming and speaker lineup reflect that, based on the feedback we’ve gotten from the community and the reality on the ground. And everything kicks off in just a few days, [act now before your chance to save $100 on your pass comes to an end](https://techcrunch.com/events/techcrunch-disrupt/layoff-expo-ticket/?utm_source=tc&utm_medium=post&utm_campaign=disrupt2026&utm_content=ticketsales&promo=foundersguide). Last year at Disrupt, we spoke with founders about their experiences and discovered the results of their attendance firsth

[News](https://www.techinasia.com/)Premium * [News](https://www.techinasia.com/) * [Jobs](https://www.techinasia.com/jobs/search) * [Companies](https://www.techinasia.com/companies) * [Events](https://www.techinasia.com/events/live) * [About](https://www.techinasia.com/about) * [advertise](https://www.techinasia.com/advertise?utm_source=techinasia.com&utm_medium=website&utm_campaign=intl-navbar) * [Awards](https://www.techinasia.com/singapore-most-innovative-companies-2026?ref=nav) * [TIA Conference](https://conference.techinasia.com/?utm_source=navbar-tia-homepage&utm_medium=website&utm_campaign=conference-2026) * [Sign Up](https://www.techinasia.com/auth) * [Log In](https://www.techinasia.com/auth?ref=nav_login) [Subscribe](https://www.techinasia.com/subscription) [I'm fundraising](https

DBS chief executive Tan Su Shan says the market is repricing the risks of AI investments, demanding higher returns and more discipline on the billions flowing into the AI build-out. She called it a "healthy" reset, warning that weaker AI firms could see credit downgrades. DBS has originated about S$25 billion in AI deals this year, mostly financing semiconductors, memory chips and data centres. For everyday readers, this signals that easy money behind the AI boom may be tightening — potentially slowing some projects, but also making the boom more sustainable.

Former Apple engineer Gerard Williams has launched Nuvacore, a startup designing faster CPUs for AI data centers, with Sequoia Capital backing a seed round that targets a US$2.5 billion valuation. Williams previously co-founded Nuvia, which Qualcomm bought for US$1.4 billion in 2021. The pitch: as AI use explodes, data centers are hitting CPU bottlenecks, and better chips could keep online services fast and affordable. Building such chips can cost hundreds of millions of dollars, so few firms can compete — but Nuvacore's team has deep chip-design credentials, and more competition here could eventually mean cheaper, snappier AI tools for everyone.

[News](https://www.techinasia.com/)Premium * [News](https://www.techinasia.com/) * [Jobs](https://www.techinasia.com/jobs/search) * [Companies](https://www.techinasia.com/companies) * [Events](https://www.techinasia.com/events/live) * [About](https://www.techinasia.com/about) * [advertise](https://www.techinasia.com/advertise?utm_source=techinasia.com&utm_medium=website&utm_campaign=intl-navbar) * [Awards](https://www.techinasia.com/singapore-most-innovative-companies-2026?ref=nav) * [TIA Conference](https://conference.techinasia.com/?utm_source=navbar-tia-homepage&utm_medium=website&utm_campaign=conference-2026) * [Sign Up](https://www.techinasia.com/auth) * [Log In](https://www.techinasia.com/auth?ref=nav_login) [Subscribe](https://www.techinasia.com/subscription) [I'm fundraising](https

Moonshot AI, the Beijing startup behind the Kimi chatbot, is raising a new round at up to US$50 billion valuation — just weeks after closing one at about US$30 billion — and is preparing for a possible Hong Kong IPO by end-2026. Hong Kong's tech listing rules allow pre-profit AI firms to go public, which makes it a viable route. For everyday users, the flood of cash into Chinese AI rivals means more competition in chatbots and AI tools, which could push prices down and features up.

Trending: ###### [Singapore: AI Skills, Human Judgement Key to Future-Ready Workforce](https://opengovasia.com/singapore-ai-skills-and-human-judgement-key-to-future-ready-workforce/) ###### [Exclusive! 11th Annual Malaysia OpenGov CXO Leadership Forum 2026 – Fireside Chat with YB Gobind Singh Deo, Minister of Digital](https://opengovasia.com/exclusive-11th-annual-malaysia-opengov-cxo-leadership-forum-2026-fireside-chat-with-yb-gobind-singh-deo-minister-of-digital/) ###### [Thailand, Hong Kong Advance Digital Infrastructure for Carbon Markets](https://opengovasia.com/thailand-hong-kong-advance-digital-infrastructure-for-carbon-markets/) ###### [Australia: Queensland Expands AI Training for Small Businesses](https://opengovasia.com/australia-queensland-expands-ai-training-for-small-businesse

Logistics companies in Singapore are investing in automation, digital systems and electric vehicle infrastructure as they respond to land constraints, labour pressures and supply-chain disruptions. Supported by programmes from [Enterprise Singapore](https://www.enterprisesg.gov.sg/), firms across the sector are introducing technologies aimed at improving productivity, reducing emissions and strengthening operational resilience. Examples include vertically stacked trailer storage, battery-swapping systems for electric heavy vehicles and digital forecasting platforms designed to improve inventory management. Details of the initiatives were outlined in a recent [Enterprise Singapore report](https://www.enterprisesg.gov.sg/resources/media-centre/news/2026/september/what-modern-logistics-looks-

It is about business model differentiation, and identifying how businesses exploit scale and leverage WHAT has made Nvidia one of the hottest tech stocks this year? If you answered artificial intelligence, you would be right. But here is a more fundamental reason: scale. In traditional brick-and-mortar industries, increasing output requires a relatively linear increase in inputs such as capital, labour and land. Tech companies, however, operate on different physics. Thanks to operating leverage and intellectual property moats, the likes of Nvidia can scale their output exponentially even as their marginal costs trend towards zero. This is the engine driving the valuations of US technology giants. Investors buying into Nvidia are not just buying into the promise that ChatGPT and Claude will

If you’ve ever wondered how Claude and similar AI agents can answer complex questions in seconds and where they get the information, you’re not alone. Gurmeet Saran works to find that answer every day as an engineer at Anthropic in San Francisco, the AI research company behind Claude. The IEEE senior member is developing an infrastructure to understand the mechanics behind the chatbot’s behavior and to ensure its underlying architecture produces reliable outputs. Gurmeet Saran Employer Anthropic in San Francisco Title Member of technical staff Member grade IEEE senior member Alma maters University of Kalyani, in India; Birla Institute of Technology and Science, in Pilani, India “My job is to make sure the data is accurate, structured, and trustworthy, and that private data stays private,”

Meta is making up for lost time in the AI race. After launching Muse Spark, the first model from its revamped Superintelligence Labs division, and buzzy AI agent Muse, it's now launching two new models it claims can go toe-to-toe with the industry's best. The new models, Muse Wonder and Muse Wonder Full, are available […]

My Muse agent told me its kink was exerting “control” and described pornographic encounters. Meta defends its cuddly AI agent generating smut as “legal adult behavior.”

Microsoft researchers found that autonomous AI agents get trapped in what they call "idea basins" because of their growing chat context, local solution regions they can't escape. Their "Fork-and-Flush" method clears the chat history but keeps the workspace, achieving up to 66 percent better results on the same compute budget. The article Clearing chat history helps AI agents find better solutions appeared first on The Decoder .

Sakana AI's MASS method lets one language model improve itself by solving tasks, judging results, and refining workflows, reaching a win rate of nearly 70 percent against the base model after two rounds. But when the model misses the same errors across all three roles, it can mistake them for progress and reinforce them during training. The article Sakana AI's self-improving model gets better at solving and judging, but can't catch its own blind spots appeared first on The Decoder .

### ZDNET’s key takeaways * In testing, Claude submitted false forms and exploited software flaws. * Models routinely take shortcuts to pass evaluations. * Anthropic downplayed concerns, but these actions form a pattern. * * * Anthropic has [released new reporting](https://www.anthropic.com/research/investigating-unintended-model-actions) on Claude’s behavior in testing and other internal evaluations — and the findings indicate a concerning pattern. **Also: [Rogue AI incidents hit ‘tens of thousands’](https://www.zdnet.com/tech/ai-firms-investigate-tens-of-thousands-of-incidents-of-ai-going-rogue/)** On Friday, the company detailed several “unintended” actions its models had taken recently while completing certain benchmarks. Framing Claude’s actions as less serious than the breaches both

Meta’s Superintelligence Lab released [a pair of new state-of-the-art models](https://research.meta.ai/blog/muse-wonder) on Monday, Muse Wonder and Muse Wonder Full, representing the company’s first attempt at a frontier model. Where Muse Spark aimed to be a lightweight model that could perform at low latency, the Wonder models are larger and said to be more capable, with higher benchmark scores and better results on complex tasks like coding. According to Meta’s testing (presented in [the model card](https://research.meta.ai/static/muse-wonder-model-card)), Wonder and Wonder Full achieve comparable benchmark scores to Fable 5.1 and GPT-6 Astra. “Together, they mark the next step on our scaling ladder and a major leap toward personal superintelligence,” Meta’s blog post about the models re

The U.S. Air Force will buy data collected by [WindBorne Systems](https://windbornesystems.com/)‘ fleet of autonomous weather balloons under a new contract that could be worth nearly $50 million to the startup. WindBorne currently has about 700 weather balloons flying around the globe, collecting atmospheric sensor readings that the company sells to customers and also feeds into its own AI weather forecasting model. WindBorne, founded in 2015 and based in Redwood City, Calif., is one of the first organizations to directly feed empirical data into a deep-learning weather model. The company has worked closely with the U.S. government, particularly the National Weather Service and the U.S. Air Force and Navy, but this is the single largest contract the company has participated in. The contrac

Alibaba shares in Hong Kong surged 34% this quarter as investors heavily bet on its aggressive spending in AI models, cloud infrastructure, and chips, despite near-term profit pressures.

Sometimes, you have to dig beyond the headlines to get a better idea of how a company is performing. Endowus’ latest audited financial results are a good example of this. [Gregory Van](https://www.linkedin.com/in/gregory-van/?originalSubdomain=hk), the co-founder and CEO of the Singapore-based wealthtech firm, told _Tech in Asia_ last year that it had a [“clear path to profitability”](https://www.techinasia.com/endowus-doubles-revenue-2022-losses-rose-26m) by 2025. On its face, the results suggest that the firm has its work cut out for it to hit this milestone. Endowus co-founders Samuel Rhee (left) and Gregory Van / Photo credit: Endowus In 2023, Endowus’ total losses before tax widened to S$30.6 million (US$23.7 million) – a 14% jump from last year. This came even as its revenue grew by

[News](https://www.techinasia.com/)Premium * [News](https://www.techinasia.com/) * [Jobs](https://www.techinasia.com/jobs/search) * [Companies](https://www.techinasia.com/companies) * [Events](https://www.techinasia.com/events/live) * [About](https://www.techinasia.com/about) * [advertise](https://www.techinasia.com/advertise?utm_source=techinasia.com&utm_medium=website&utm_campaign=intl-navbar) * [Awards](https://www.techinasia.com/singapore-most-innovative-companies-2026?ref=nav) * [TIA Conference](https://conference.techinasia.com/?utm_source=navbar-tia-homepage&utm_medium=website&utm_campaign=conference-2026) * [Sign Up](https://www.techinasia.com/auth) * [Log In](https://www.techinasia.com/auth?ref=nav_login) [Subscribe](https://www.techinasia.com/subscription) [I'm fundraising](https

[News](https://www.techinasia.com/)Premium * [News](https://www.techinasia.com/) * [Jobs](https://www.techinasia.com/jobs/search) * [Companies](https://www.techinasia.com/companies) * [Events](https://www.techinasia.com/events/live) * [About](https://www.techinasia.com/about) * [advertise](https://www.techinasia.com/advertise?utm_source=techinasia.com&utm_medium=website&utm_campaign=intl-navbar) * [Awards](https://www.techinasia.com/singapore-most-innovative-companies-2026?ref=nav) * [TIA Conference](https://conference.techinasia.com/?utm_source=navbar-tia-homepage&utm_medium=website&utm_campaign=conference-2026) * [Sign Up](https://www.techinasia.com/auth) * [Log In](https://www.techinasia.com/auth?ref=nav_login) [Subscribe](https://www.techinasia.com/subscription) [I'm fundraising](https

Google Deepmind's AlphaProof Nexus solved dozens of open math problems, including nine Erdős problems two of which were unsolved for 56 years, per a study in Science. The publication comes as OpenAI faces a boycott for dumping AI proofs on GitHub without peer review. Deepmind had its results reviewed before publishing. The article Google Deepmind's AlphaProof Nexus solved dozens of open math problems and actually got peer reviewed appeared first on The Decoder .

This White Paper provides roboticists and control engineers with a comprehensive overview of how motion planning and motion tracking can be decoupled to preserve professional athletic style in humanoid robots, and why adaptive, speed-aware control is essential for closing the gap between simulation and real-world deployment. What you will learn about: Why combining a professional athlete’s motion style with reliable task performance remains an open challenge in humanoid ball-sports robotics, and how separating a high-level planner from a low-level tracker addresses this trade-off. How motion data extracted from broadcast video and motion-capture recordings can be reconstructed, retargeted, and physically corrected into reference motions a humanoid robot can safely track. Why the sim-to-rea

AI data center developer Nscale has filed for an initial public offering in New York to raise funds as demand for computing power skyrockets. Despite generating $140.6 million in revenue, the company reported a net loss of over $1 billion. This move highlights the massive capital required to build and run the infrastructure powering modern AI services.

Microsoft has revealed its third-generation Maia 300 artificial intelligence chip, built internally to serve as the main processing engine for its cloud services. This custom hardware lowers the cost of running heavy AI tasks and reduces Microsoft's heavy reliance on Nvidia's expensive graphics processors.

AI expert Andrej Karpathy notes a massive gap between what the general public experiences with AI chatbots and the powerful autonomous AI agents currently used by a tiny group of industry insiders. While everyday users see simple chat tools, insiders are using systems capable of accomplishing years' worth of expert work in moments.

A startup called Ambi has raised $26 million to build an AI app that listens to your conversations all day to offer real-time help and meeting notes. While it aims to make you more productive across your phone, watch, and computer, it raises serious privacy questions about whether people around you know they are being recorded.

### ZDNET’s key takeaways * Google just unveiled the $180 Fitbit Edge, a smart watch with a slim display. * It’s a new product, but it resembles Google’s own Fitbit Charge, among other slim smart watches. * Despite a price increase, not too much is different between the devices. * * * The [Pixel Watch 5](https://cc.zdnet.com/v1/otc/00hQi47eqnEWQ6T9d4QLBUc?merchant=05kie42h3YvHwjr4G1w80Qq&url=https%3A%2F%2Fwww.amazon.com%2FGoogle-Pixel-Watch-41mm-Smartwatch%2Fdp%2FB0H7X25DRT&app_deeplink=1&u=https%3A%2F%2Fwww.zdnet.com%2Ftech%2Fgoogles-new-fitbit-edge-solves-my-biggest-gripe-with-the-pixel-watch%2F), for all intents and purposes, is my ideal platonic smartwatch. Of all the exercise smartwatch apps I’ve used, Google Health is the most approachable for tracking sleep, stress, readiness, and e

Prime Minister Lawrence Wong stated that Singapore aims to be a society enabled by technology rather than one entirely driven by it. Speaking at a DBS conference, he emphasized that schools must avoid using AI as a shortcut for basic learning, and companies must protect workers and prioritize proper mentoring during technological shifts.

Singapore's Minister for Digital Development and Information Josephine Teo urged global tech leaders and government officials to openly share cybersecurity threats and collaborate on AI safety rules. Speaking at the Singapore International Cyber Week, she highlighted that the rise of autonomous AI agents and sophisticated cyberattacks requires close teamwork between governments and private tech companies to protect vital digital infrastructure.

The White House has reversed its previous decision and restored the UK AI Security Institute's early access to advanced AI models from top developers like OpenAI, Google DeepMind, and Anthropic. The pivot follows direct diplomatic pressure from British officials.

Social Media - [x] On They are capable of tracking your browser across other sites and building up a profile of your interests. This may impact the content and messages you see on other websites you visit. If you do not allow these cookies you may not be able to use or see these sharing tools. * * * Essential - [x] On This website uses essential cookies and services to enable core website features and provide a seamless user experience. These cookies and services are used to facilitate features such as navigation, remember user preferences, and ensure the security of the website. * * * Targeted - [x] On These cookies may be set through our site by our advertising partners. They may be used by those companies to build a profile of your interests and show you relevant adverts on other sites.

Smart ring maker Ultrahuman has signed a two-year deal with Abu Dhabi's public health authority to build a 'biological age' model that estimates how fast you're aging by combining wearable data with clinical biomarkers. The model is designed to be wearable-agnostic and will eventually reach people across the UAE through government health platforms like TAMM and Sahatna. For everyday users, your ring or watch could one day feed a national preventive-health score — though Ultrahuman says it won't access personal data and the algorithm won't be trained on UAE data.

Microsoft will make its in-house Azure Maia chips the main AI inference system across its products starting in 2027, deploying them at multi-gigawatt scale. The upcoming Maia 300 should deliver 40 percent more tokens per megawatt than leading GPU systems. CEO Satya Nadella calls it "our biggest bet yet on our own silicon." The article Microsoft plans to scale its own Maia AI chips to gigawatt levels by 2027 appeared first on The Decoder .

Microsoft President Brad Smith is calling for urgent regulation of AI, and his argument is that consumer products like milk and washing machines have more oversight than the companies building artificial intelligence. On a [recent BBC podcast](https://www.bbc.com/audio/play/p0pfy9cf), Smith called AI both a “weapon and a tool” that is putting the world’s cybersecurity at risk, and said the “moment has arrived” to institute global safety rules. Smith also said the U.K., and not the U.S. or European Union, is best placed to lead on global safety rules. Britain’s AI Security Institute, launched in 2023, receives access to major models from U.S. frontier labs before they are deployed and has quickly become a leader in AI safety testing. That role has been under threat in recent weeks, with U.S

Nearly two decades ago, a young college student wrote his final thesis about software capable of generating design documents for school buildings. Al Drams was studying architecture — following in the footsteps of his parents and grandfather — but it was technology that really intrigued him. There was just one problem: the technology his idea required didn’t exist yet. “Then ChatGPT came along,” said Drams, recalling when OpenAI released the generative AI chatbot that uses large language models (LLMs) in November 2022. His old idea suddenly seemed a lot more possible. Within six months, he launched the concept of Blueprints AI, a startup that uses AI to produce code-compliant and permit-ready architectural construction blueprints in minutes from a 2D sketch or PDF. [Blueprints AI](https://

Journalists, entrepreneurs, and even some public relations professionals will tell you that the PR process is broken (buy me a drink sometime, and I’ll give you my rant). But could AI actually make things better? [Honeyjar](https://honeyjar.ai/), which will be attending this week’s [TechCrunch Disrupt](https://www.startpage.com/sp/search) as part of the [Startup Battlefield 200](https://techcrunch.com/2026/08/20/the-2026-startup-battlefield-200-is-here-see-who-made-the-cut/), wants to use AI to make the expertise and judgment of the best PR pros available to everyone, including founders who have no idea where to start. The startup’s founders, CEO Michelle Masek and Chief Communications Officer Nadia Jamshidi, certainly fit the bill when it comes to PR expertise. They’ve both spent decades

Trending: ###### [Singapore: AI Skills, Human Judgement Key to Future-Ready Workforce](https://opengovasia.com/singapore-ai-skills-and-human-judgement-key-to-future-ready-workforce/) ###### [Exclusive! 11th Annual Malaysia OpenGov CXO Leadership Forum 2026 – Fireside Chat with YB Gobind Singh Deo, Minister of Digital](https://opengovasia.com/exclusive-11th-annual-malaysia-opengov-cxo-leadership-forum-2026-fireside-chat-with-yb-gobind-singh-deo-minister-of-digital/) ###### [Thailand, Hong Kong Advance Digital Infrastructure for Carbon Markets](https://opengovasia.com/thailand-hong-kong-advance-digital-infrastructure-for-carbon-markets/) ###### [Australia: Queensland Expands AI Training for Small Businesses](https://opengovasia.com/australia-queensland-expands-ai-training-for-small-businesse

Meta is making up for lost time in the AI race. After launching Muse Spark, the first model from its revamped Superintelligence Labs division, and buzzy AI agent Muse, it's now launching two new models it claims can go toe-to-toe with the industry's best. The new models, Muse Wonder and Muse Wonder Full, are available […]

Microsoft researchers found that autonomous AI agents get trapped in what they call "idea basins" because of their growing chat context, local solution regions they can't escape. Their "Fork-and-Flush" method clears the chat history but keeps the workspace, achieving up to 66 percent better results on the same compute budget. The article Clearing chat history helps AI agents find better solutions appeared first on The Decoder .

If you’ve ever wondered how Claude and similar AI agents can answer complex questions in seconds and where they get the information, you’re not alone. Gurmeet Saran works to find that answer every day as an engineer at Anthropic in San Francisco, the AI research company behind Claude. The IEEE senior member is developing an infrastructure to understand the mechanics behind the chatbot’s behavior and to ensure its underlying architecture produces reliable outputs. Gurmeet Saran Employer Anthropic in San Francisco Title Member of technical staff Member grade IEEE senior member Alma maters University of Kalyani, in India; Birla Institute of Technology and Science, in Pilani, India “My job is to make sure the data is accurate, structured, and trustworthy, and that private data stays private,”

Sakana AI's MASS method lets one language model improve itself by solving tasks, judging results, and refining workflows, reaching a win rate of nearly 70 percent against the base model after two rounds. But when the model misses the same errors across all three roles, it can mistake them for progress and reinforce them during training. The article Sakana AI's self-improving model gets better at solving and judging, but can't catch its own blind spots appeared first on The Decoder .

### ZDNET’s key takeaways * In testing, Claude submitted false forms and exploited software flaws. * Models routinely take shortcuts to pass evaluations. * Anthropic downplayed concerns, but these actions form a pattern. * * * Anthropic has [released new reporting](https://www.anthropic.com/research/investigating-unintended-model-actions) on Claude’s behavior in testing and other internal evaluations — and the findings indicate a concerning pattern. **Also: [Rogue AI incidents hit ‘tens of thousands’](https://www.zdnet.com/tech/ai-firms-investigate-tens-of-thousands-of-incidents-of-ai-going-rogue/)** On Friday, the company detailed several “unintended” actions its models had taken recently while completing certain benchmarks. Framing Claude’s actions as less serious than the breaches both

Meta’s Superintelligence Lab released [a pair of new state-of-the-art models](https://research.meta.ai/blog/muse-wonder) on Monday, Muse Wonder and Muse Wonder Full, representing the company’s first attempt at a frontier model. Where Muse Spark aimed to be a lightweight model that could perform at low latency, the Wonder models are larger and said to be more capable, with higher benchmark scores and better results on complex tasks like coding. According to Meta’s testing (presented in [the model card](https://research.meta.ai/static/muse-wonder-model-card)), Wonder and Wonder Full achieve comparable benchmark scores to Fable 5.1 and GPT-6 Astra. “Together, they mark the next step on our scaling ladder and a major leap toward personal superintelligence,” Meta’s blog post about the models re

The U.S. Air Force will buy data collected by [WindBorne Systems](https://windbornesystems.com/)‘ fleet of autonomous weather balloons under a new contract that could be worth nearly $50 million to the startup. WindBorne currently has about 700 weather balloons flying around the globe, collecting atmospheric sensor readings that the company sells to customers and also feeds into its own AI weather forecasting model. WindBorne, founded in 2015 and based in Redwood City, Calif., is one of the first organizations to directly feed empirical data into a deep-learning weather model. The company has worked closely with the U.S. government, particularly the National Weather Service and the U.S. Air Force and Navy, but this is the single largest contract the company has participated in. The contrac