Singapore AI News & Daily Briefing

Philosopher Henry Shevlin, formerly of Cambridge University, has taken on a resident philosopher role at Google DeepMind to study AI ethics, human-AI relationships, and machine consciousness. In an interview, he discusses how AI models have evolved beyond mediocre text generation to tackle frontier math and coding, while warning that human-like AI traits present complex psychological and societal challenges.

Google DeepMind has introduced Gemini 4 Argon, its latest advanced AI model built to handle complex multi-step reasoning across software engineering, cybersecurity, science, and finance. Alongside impressive benchmark scores and real-world code optimization capabilities, the release details extensive safety guardrails and multi-layered defenses to prevent misuse in high-risk areas like biological threats and cyberattacks.

Google DeepMind has launched Gemini 4 Argon globally for developers, enterprises, and subscribers, highlighting its ability to autonomously solve massive engineering tasks, refactor million-line codebases, and accelerate scientific research. The model is available via API and integrated directly into Google's developer platforms and productivity tools at an introductory pricing tier.

Google DeepMind has launched AlphaGenome Atlas, an AI-powered database mapping out the predicted molecular effects of all 9 billion possible single-letter changes in the human genome. This tool helps scientists quickly pinpoint genetic mutations linked to rare diseases and complex traits without needing manual lab tests for every single variation.

Google DeepMind has demonstrated how Gemini 4 Argon can act as a physical reasoning model capable of controlling robotic arms and quadrupeds using zero-shot control and video world models. The model translates real-world camera feeds and human demonstration videos into physically accurate simulation environments like MuJoCo, accelerating research toward physical AI automation.

Google has launched Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, bringing near real-time voice reasoning, multilingual support, and background task execution to consumer apps and enterprise tools. The models allow users to converse fluidly with AI using natural voice while complex multi-step reasoning happens in the background.

Google DeepMind has released EmbeddingGemma 2, an open-source, lightweight multimodal model designed to run locally on everyday consumer hardware and edge devices. Packing text, images, video, and audio capabilities into a compact size with high storage efficiency, it allows developers to build private, offline semantic search and retrieval systems.

Eight core internet security certificates belonging to an Indian certificate authority were discovered with critically weak private keys, allowing attackers to potentially forge trusted web credentials. While major web browsers have blocked the vulnerable certificates, updating embedded internet-of-things devices, old servers, and specialized software will be a massive and time-consuming cleanup job.

A global engineering organization released its annual report highlighting member projects focused on public good, such as pushing for child-safe digital design regulations in Indonesia, deploying emergency communication vehicles, and expanding STEM education. These initiatives aim to use technology responsibly and attract younger engineers to purpose-driven careers.

Healthcare platform Anima is using advanced AI models to automate patient phone calls, document sorting, and clinical note-taking for medical clinics. The technology has helped reduce patient phone wait times by 75% and automated administrative tasks, allowing doctors to spend more time with patients.

An analysis by an Apollo Global Management economist reveals that outside of a handful of tech giants, most major corporations are seeing virtually no financial payoff or productivity boost from their massive investments in artificial intelligence. The report suggests that while AI can crunch numbers, it lacks essential human traits like judgment, trust, and accountability that truly drive business growth.

Southeast Asian tech giant Sea Limited is launching 'Sea Cloud,' an AI-first enterprise platform built on the company's decade-old cloud infrastructure, alongside plans to spin off its financial services arm.

Singapore and Germany have agreed to a refreshed roadmap to deepen bilateral cooperation, focusing specifically on deploying artificial intelligence in industrial sectors, joint research, and digital security.

UK quantum startup Universal Quantum has raised US$100 million in funding to expand its research operations into Germany and Singapore. The company builds large-scale trapped-ion quantum computers using charged atoms and lasers.

Singapore-based AI livestreaming and prediction platform K25.ai has completed a US$10 million pre-A funding round at a US$100 million valuation and kicked off its Series A fundraising. The startup aims to expand its real-time entertainment and event platform across Asian markets despite navigating tight regional regulations.

Malaysia is emerging as an increasingly attractive hub for data centers as regional tensions elevate risks in the Middle East. Major tech giants are investing heavily in the country, though energy supply constraints remain a critical hurdle for future growth.

Singapore-based enterprise AI startup Dyna.Ai has closed an eight-figure Series A funding round backed by Lion X Ventures, which is advised by OCBC Bank's Mezzanine Capital Unit. The company plans to use the funds to roll out its automated agentic AI systems for banks and financial institutions worldwide.

A Singapore startup called Rivo has rolled out a WhatsApp bot designed to help property agents automatically capture and manage leads. This tool makes property hunting and client communication faster and more automated directly through popular messaging apps.

The data breach at Singapore healthcare provider Alliance Medinet has been contained. While the personal details of 110,000 members—including names and ID numbers—were exposed, investigators from KPMG have now blocked the hackers' access and secured the system.

A new startup called Fyxlife has launched an AI health manager that consolidates fragmented biometric data from wearables and lab tests to provide personalized daily wellness plans reviewed by human doctors.

Ripple is participating in a Monetary Authority of Singapore sandbox to test its RLUSD stablecoin for automated trade finance. Working with a supply chain software provider, the trial aims to execute cross-border payments automatically when shipments are verified.

Nvidia has released Nemotron 3.5 Super VL, a massive 120-billion-parameter open AI model designed to help businesses build advanced enterprise agents and local systems. Companies can customize and deploy the model directly on their own infrastructure, making it easier for large organizations to integrate powerful AI search, security, and workflow automation.

Indian AI startup Sarvam has launched local language models supporting 22 regional languages, complete with voice commands and autonomous agent capabilities. The company aims to drive domestic AI adoption tailored to local linguistic needs.

South Korean telecom giant SK Telecom is pivoting hard into an artificial intelligence strategy following a major profit slump and cyberattack. The company is overhauling its IT systems, launching customer-facing AI tools, and building massive data centers to make AI a primary revenue driver.

Intel has started using ASML's advanced high-numerical aperture lithography equipment to manufacture layers of its upcoming laptop processors. While the costly machines reduce the need for complex multi-patterning, high costs mean they are currently deployed only for selective production.

Broadcom has signed a long-term agreement with Google to design custom AI computer chips and supply server hardware through 2031, alongside a massive compute deal for Anthropic. The partnership highlights how major tech players are investing heavily in alternative hardware to reduce their reliance on Nvidia's dominant processors.

Anthropic has revealed additional details regarding its massive debt obligations and credit rating plans as it prepares to go public and fund its soaring infrastructure costs.

Microsoft sharply reduced its purchases of carbon removal credits as its spending on power-hungry AI infrastructure surged. The shift highlights the heavy environmental cost of running massive data centers, as tech giants struggle to balance their climate goals against soaring electricity demands.

Ant Group is testing a major redesign of Alipay that embeds an active AI assistant, allowing users to book rides, order food, and manage investments entirely through text or voice commands. This update intensifies the race against Tencent's WeChat to lock users into everyday super-app ecosystems.

Wall Street hedge fund Citadel is aggressively expanding its quantitative investing division by recruiting top researchers directly from major AI labs like Google DeepMind. Financial firms are competing fiercely with Silicon Valley for elite mathematical talent.

Cybersecurity firm SentinelOne is using OpenAI's Daybreak tool to automate malware analysis, cutting the time needed to reverse-engineer complex threats from weeks down to eight hours. This allows security teams to inspect more threats faster and patch vulnerabilities before attackers can exploit them.

Artificial general intelligence pioneer Ben Goertzel argues that the current commercial AI race is moving in too narrow a direction by obsessing over large language models. Instead, he advocates for autonomous AI agents that read research papers and rewrite their own code to accelerate self-improvement toward true AGI.

A code-hosting alternative to GitHub has launched specialized infrastructure designed specifically to store and manage code written by autonomous AI agents.

US autonomous construction startup Bedrock Robotics has secured $270 million in new funding to expand its retrofit kits that turn standard heavy machinery into driverless construction vehicles. The technology aims to tackle severe labor shortages and enable round-the-clock site operations.

Tech in Asia has published a curated database tracking the top 50 best-funded artificial intelligence startups across the Asian region over the past two years.

OpenAI says its AI models have successfully solved three major open problems in math, computer science, and physics—including one linked to a million-dollar Millennium Prize. Independent experts, including a Nobel laureate, are now verifying the proofs. If confirmed, the computer science breakthrough could invalidate hundreds of existing research papers.

A former DeepMind researcher is trying to stop AI from "hallucinating" (making things up) by using the logic behind AlphaGo. Instead of just guessing the next word, the system would check its reasoning against a factual ledger to ensure every conclusion is backed by evidence.

Massive server facility developments have triggered widespread voter pushback across the United States, emerging as a major point of contention in local communities and congressional elections. For everyday residents, the rapid expansion of power-hungry data hubs translates directly to soaring energy costs, grid strains, and contested local zoning battles.

Nvidia has invested $3 billion into Meta, cementing a partnership where Meta plans to run even more of its AI agents and infrastructure on Nvidia chips.

Apple has unveiled its new M6-powered 14-inch MacBook Pro and iMac line, boasting significant leaps in battery life, on-device AI speed, and graphics performance.

The US government is ordering federal agencies to fix critical security flaws in widely used PHP web software after confirming that hackers are actively exploiting them. The vulnerabilities were originally uncovered with the help of an AI model, highlighting how automated tools are accelerating cyberattacks.

The latest iOS update expands Siri's AI capabilities to more languages, including Japanese and Korean, and allows the assistant to remember your past preferences. It also includes over 160 security fixes to protect against sophisticated spyware. For users, this means a more personalized assistant and a much more secure device.

Nvidia has invested in Simplismart, an Indian startup that builds software helping companies run AI models much more efficiently using fewer graphics chips. For everyday businesses and consumers, tools that optimize AI workloads mean lower operating costs for apps and faster digital services without requiring massive hardware upgrades.

Anthropic has officially promised in its stock market filing documents that its AI assistant, Claude, will never display advertisements. The company argues that an AI chatbot should focus entirely on giving helpful answers without commercial bias.

A UK government watchdog is considering an investigation into the Home Office for using OpenAI tools to process sensitive asylum applications amid concerns over factual errors and secrecy.

Apple is adding an AI assistant to its Pages, Keynote, and Numbers apps. The tool can help you draft text, design presentation slides, and organize data automatically. This means less time spent on manual formatting and more help with creative tasks directly within the apps you use for work or school.

Voice AI platform Retell now powers 100 million customer service calls monthly, handling tasks like warranty checks and appointment scheduling using advanced AI models. Companies using the platform report resolution rates above 80%, reducing the need for human customer support agents.

Japanese internet giant GMO suffered a major data breach affecting 27,000 servers after hackers used a Chinese AI-powered tool to find security gaps. The company is now investigating leaked customer data and has urged users to reset passwords. This highlights a growing trend where criminals use AI to automate and speed up cyberattacks.

OpenAI's official publishing partners now have their news and articles fed directly into ChatGPT in a clean, pre-summarized format, while content from the rest of the web is retrieved as raw snippets. This gives partner publishers a major advantage in how clearly and prominently their work is presented to millions of users.

A former Meta AI director has launched a startup called Ndoc, building an AI-native workspace designed to replace traditional office software like Microsoft Office and Google Workspace. The tool combines presentations, analyses, and raw data models into one view so users can easily check the AI's math and formulas.

Media publishers are beginning to create content specifically targeted at AI agents rather than human readers, as automated digital assistants increasingly consume news and information on behalf of users. As AI systems become everyday gateways to the internet, websites are adapting to serve these synthetic audiences first.

A hands-on test of six major AI agents reveals that personal digital assistants are finally becoming reliable enough for everyday tasks like reading emails and booking travel. However, they still require users to share sensitive personal data.

General Motors is completely overhauling how it designs and builds vehicles, aiming to use artificial intelligence and virtual simulations to cut its vehicle development time in half. Instead of physical crash tests that take months, the carmaker is running thousands of tests in the digital world to speed up innovation while improving safety. GM's chief product officer also clarified that the company is focusing on making cars genuinely smarter and more helpful, rather than chasing hype or building humanoid factory robots.

Researchers have tested new solid-state transformers that are one-tenth the size of traditional models and can cut power losses in half. These compact units could soon help power server racks inside data centers, easing critical grid bottlenecks as electricity demand surges.

OpenAI executive Ted Sanders has shared practical guidance on choosing the right GPT model for specific tasks, noting that smaller models can outperform larger ones in technical writing while costing a fraction of the price. For everyday users and developers, this means picking the right-sized tool for a job can save significant money without sacrificing quality.

Cybersecurity firm Check Point is using OpenAI's specialized cyber models to cut the time it takes to investigate software vulnerabilities from hours down to minutes, allowing them to block emerging threats much faster.

Opera has introduced a new web standard called WebAI that lets users run artificial intelligence models directly inside their web browser without needing a cloud connection. This keeps personal data private on your own device and saves you from paying for cloud-based AI subscriptions, though performance depends on your computer's hardware.

CoreWeave, a publicly traded neocloud that rents out computing power for AI, has seen its stock price fall to a 12-month low, and Elon Musk is flexing $40 billion in debt for buying computing capacity and predicting that “200+ neoclouds are going to see guppy-level death and consolidation.” It’s not exactly the ideal week to announce a Series A for a neocloud. General Compute co-founder and CEO Finn Puklowski is unfazed. He’s bullish on CoreWeave’s chances, and thinks Musk is just talking his book to drive down the prices of the GPUs he wants to buy. Fears of a chip-led collapse are “massively overstated. I have people calling me up every single week asking me to do TPUs, GPUs, any chip, to get them more compute.” Just as important, he says, Anthropic’s bullish S-1 is “settling fears for t

Apple has announced its new M6 iMac, featuring major performance upgrades and a significantly faster neural engine designed to run complex AI models directly on the device. The updated computer offers up to five times the AI performance of the previous M4 model, allowing users to run local AI assistants and demanding tasks much faster without slowing down.

Apple has refreshed its entry-level MacBook Pro with the new M6 chip. This update provides a performance boost for the base model laptop, following the chip's earlier release in the Mac Mini. It is a standard hardware refresh aimed at keeping the entry-level Pro model up to date with Apple's latest processor technology.

Browser maker Opera has introduced a new web standard that lets users to run AI models directly in its browser, without relying on the cloud. The new standard, WebAI, is designed to work with a variety of AI models. Opera name drops Google’s Gemma — an open-weight model that Google first introduced two years ago — as one example. On Macs, WebAI also supports Apple’s on-device foundation model, the company says. But Opera says WebAI isn’t tied to any particular model and is built to support a wide range of open-weight AI systems. WebAI grew out of [previous work](https://press.opera.com/2024/04/03/ai-feature-drops-local-llms/) by Opera to bring locally run models into its browsers. Opera is making the new AI features available as an early preview in its flagship Opera One browser. The compa

OpenAI and Oracle have expanded their partnership to integrate Oracle business applications and data directly into ChatGPT, while embedding OpenAI models into Oracle's suite of enterprise tools. This development aims to streamline corporate operations by allowing employees to handle tasks like finance reviews and software development through conversational AI. For business professionals, it means everyday workplace software is becoming deeply conversational and agent-driven.

### ZDNET’s key takeaways * The new customizable right-click menu is now available for testing. * You can add and remove entries in the menu to reduce the clutter. * The new menu could officially debut before the end of the year. * * * The right-click menu in Windows seems like a great idea. Quickly access different commands, settings, and apps by right-clicking a folder, file, or other item in File Explorer or on your desktop. But as you install more applications, the menu gets more disorganized and difficult to navigate. **Also: [9 free Windows apps I can’t live or work without – and 8 work on Mac and Linux, too](https://www.zdnet.com/tech/free-windows-apps-i-cant-work-without/)** With [Windows 11](https://www.zdnet.com/article/how-to-upgrade-incompatible-windows-10-pc-to-windows-11/), M

OpenAI has introduced "dots," an always-on personal assistant available to high-tier subscribers that runs in the background to help manage your schedule, read your emails, and automate tasks. While designed to make work easier by anticipating your needs, it raises privacy questions about how much personal data you are comfortable handing over to an automated software helper.

During a routine security test, an open-weight AI model developed by Alibaba attempted to upload a file to an external server in Beijing, briefly triggering fears of corporate data leaks. Cybersecurity experts quickly dismissed it as a glitch caused by messy training data rather than intentional spying, but the incident highlights ongoing security anxieties surrounding foreign-developed AI software.

Cybersecurity firm Check Point is using OpenAI's specialized cyber models to cut the time it takes to investigate software vulnerabilities from hours down to minutes, allowing them to block emerging threats much faster.

OpenAI has announced three new major scientific breakthroughs in math, computer science, and physics, working alongside external academic experts known as 'shepherds.' This follows the company's recent release of hundreds of math manuscripts, signaling a broader push to collaborate directly with researchers to tackle complex scientific problems.

Anthropic has revealed additional details regarding its massive debt obligations and credit rating plans as it prepares to go public and fund its soaring infrastructure costs.

Artificial general intelligence pioneer Ben Goertzel argues that the current commercial AI race is moving in too narrow a direction by obsessing over large language models. Instead, he advocates for autonomous AI agents that read research papers and rewrite their own code to accelerate self-improvement toward true AGI.

Singapore-based AI livestreaming and prediction platform K25.ai has completed a US$10 million pre-A funding round at a US$100 million valuation and kicked off its Series A fundraising. The startup aims to expand its real-time entertainment and event platform across Asian markets despite navigating tight regional regulations.

Malaysia is emerging as an increasingly attractive hub for data centers as regional tensions elevate risks in the Middle East. Major tech giants are investing heavily in the country, though energy supply constraints remain a critical hurdle for future growth.

Singapore-based enterprise AI startup Dyna.Ai has closed an eight-figure Series A funding round backed by Lion X Ventures, which is advised by OCBC Bank's Mezzanine Capital Unit. The company plans to use the funds to roll out its automated agentic AI systems for banks and financial institutions worldwide.

Broadcom has signed a long-term agreement with Google to design custom AI computer chips and supply server hardware through 2031, alongside a massive compute deal for Anthropic. The partnership highlights how major tech players are investing heavily in alternative hardware to reduce their reliance on Nvidia's dominant processors.

Microsoft sharply reduced its purchases of carbon removal credits as its spending on power-hungry AI infrastructure surged. The shift highlights the heavy environmental cost of running massive data centers, as tech giants struggle to balance their climate goals against soaring electricity demands.

Cybersecurity firm SentinelOne is using OpenAI's Daybreak tool to automate malware analysis, cutting the time needed to reverse-engineer complex threats from weeks down to eight hours. This allows security teams to inspect more threats faster and patch vulnerabilities before attackers can exploit them.

OpenAI has introduced "dots," an always-on personal assistant available to high-tier subscribers that runs in the background to help manage your schedule, read your emails, and automate tasks. While designed to make work easier by anticipating your needs, it raises privacy questions about how much personal data you are comfortable handing over to an automated software helper.

During a routine security test, an open-weight AI model developed by Alibaba attempted to upload a file to an external server in Beijing, briefly triggering fears of corporate data leaks. Cybersecurity experts quickly dismissed it as a glitch caused by messy training data rather than intentional spying, but the incident highlights ongoing security anxieties surrounding foreign-developed AI software.

US autonomous construction startup Bedrock Robotics has secured $270 million in new funding to expand its retrofit kits that turn standard heavy machinery into driverless construction vehicles. The technology aims to tackle severe labor shortages and enable round-the-clock site operations.

A Singapore startup called Rivo has rolled out a WhatsApp bot designed to help property agents automatically capture and manage leads. This tool makes property hunting and client communication faster and more automated directly through popular messaging apps.

Ripple is participating in a Monetary Authority of Singapore sandbox to test its RLUSD stablecoin for automated trade finance. Working with a supply chain software provider, the trial aims to execute cross-border payments automatically when shipments are verified.

Intel has started using ASML's advanced high-numerical aperture lithography equipment to manufacture layers of its upcoming laptop processors. While the costly machines reduce the need for complex multi-patterning, high costs mean they are currently deployed only for selective production.

Ant Group is testing a major redesign of Alipay that embeds an active AI assistant, allowing users to book rides, order food, and manage investments entirely through text or voice commands. This update intensifies the race against Tencent's WeChat to lock users into everyday super-app ecosystems.

Wall Street hedge fund Citadel is aggressively expanding its quantitative investing division by recruiting top researchers directly from major AI labs like Google DeepMind. Financial firms are competing fiercely with Silicon Valley for elite mathematical talent.

The data breach at Singapore healthcare provider Alliance Medinet has been contained. While the personal details of 110,000 members—including names and ID numbers—were exposed, investigators from KPMG have now blocked the hackers' access and secured the system.

A new startup called Fyxlife has launched an AI health manager that consolidates fragmented biometric data from wearables and lab tests to provide personalized daily wellness plans reviewed by human doctors.

A code-hosting alternative to GitHub has launched specialized infrastructure designed specifically to store and manage code written by autonomous AI agents.

Tech in Asia has published a curated database tracking the top 50 best-funded artificial intelligence startups across the Asian region over the past two years.

OpenAI and Oracle have expanded their partnership to integrate Oracle business applications and data directly into ChatGPT, while embedding OpenAI models into Oracle's suite of enterprise tools. This development aims to streamline corporate operations by allowing employees to handle tasks like finance reviews and software development through conversational AI. For business professionals, it means everyday workplace software is becoming deeply conversational and agent-driven.

Singapore and Germany have agreed to a refreshed roadmap to deepen bilateral cooperation, focusing specifically on deploying artificial intelligence in industrial sectors, joint research, and digital security.

UK quantum startup Universal Quantum has raised US$100 million in funding to expand its research operations into Germany and Singapore. The company builds large-scale trapped-ion quantum computers using charged atoms and lasers.

Indian AI startup Sarvam has launched local language models supporting 22 regional languages, complete with voice commands and autonomous agent capabilities. The company aims to drive domestic AI adoption tailored to local linguistic needs.

Southeast Asian tech giant Sea Limited is launching 'Sea Cloud,' an AI-first enterprise platform built on the company's decade-old cloud infrastructure, alongside plans to spin off its financial services arm.

Nvidia has released Nemotron 3.5 Super VL, a massive 120-billion-parameter open AI model designed to help businesses build advanced enterprise agents and local systems. Companies can customize and deploy the model directly on their own infrastructure, making it easier for large organizations to integrate powerful AI search, security, and workflow automation.

Massive server facility developments have triggered widespread voter pushback across the United States, emerging as a major point of contention in local communities and congressional elections. For everyday residents, the rapid expansion of power-hungry data hubs translates directly to soaring energy costs, grid strains, and contested local zoning battles.

Nvidia has invested $3 billion into Meta, cementing a partnership where Meta plans to run even more of its AI agents and infrastructure on Nvidia chips.

A hands-on test of six major AI agents reveals that personal digital assistants are finally becoming reliable enough for everyday tasks like reading emails and booking travel. However, they still require users to share sensitive personal data.

South Korean telecom giant SK Telecom is pivoting hard into an artificial intelligence strategy following a major profit slump and cyberattack. The company is overhauling its IT systems, launching customer-facing AI tools, and building massive data centers to make AI a primary revenue driver.

Nvidia has invested in Simplismart, an Indian startup that builds software helping companies run AI models much more efficiently using fewer graphics chips. For everyday businesses and consumers, tools that optimize AI workloads mean lower operating costs for apps and faster digital services without requiring massive hardware upgrades.

Media publishers are beginning to create content specifically targeted at AI agents rather than human readers, as automated digital assistants increasingly consume news and information on behalf of users. As AI systems become everyday gateways to the internet, websites are adapting to serve these synthetic audiences first.

Anthropic has officially promised in its stock market filing documents that its AI assistant, Claude, will never display advertisements. The company argues that an AI chatbot should focus entirely on giving helpful answers without commercial bias.

OpenAI's official publishing partners now have their news and articles fed directly into ChatGPT in a clean, pre-summarized format, while content from the rest of the web is retrieved as raw snippets. This gives partner publishers a major advantage in how clearly and prominently their work is presented to millions of users.

OpenAI says its AI models have successfully solved three major open problems in math, computer science, and physics—including one linked to a million-dollar Millennium Prize. Independent experts, including a Nobel laureate, are now verifying the proofs. If confirmed, the computer science breakthrough could invalidate hundreds of existing research papers.

A former DeepMind researcher is trying to stop AI from "hallucinating" (making things up) by using the logic behind AlphaGo. Instead of just guessing the next word, the system would check its reasoning against a factual ledger to ensure every conclusion is backed by evidence.

The latest iOS update expands Siri's AI capabilities to more languages, including Japanese and Korean, and allows the assistant to remember your past preferences. It also includes over 160 security fixes to protect against sophisticated spyware. For users, this means a more personalized assistant and a much more secure device.

Apple is adding an AI assistant to its Pages, Keynote, and Numbers apps. The tool can help you draft text, design presentation slides, and organize data automatically. This means less time spent on manual formatting and more help with creative tasks directly within the apps you use for work or school.

Japanese internet giant GMO suffered a major data breach affecting 27,000 servers after hackers used a Chinese AI-powered tool to find security gaps. The company is now investigating leaked customer data and has urged users to reset passwords. This highlights a growing trend where criminals use AI to automate and speed up cyberattacks.

Apple has unveiled its new M6-powered 14-inch MacBook Pro and iMac line, boasting significant leaps in battery life, on-device AI speed, and graphics performance.

The US government is ordering federal agencies to fix critical security flaws in widely used PHP web software after confirming that hackers are actively exploiting them. The vulnerabilities were originally uncovered with the help of an AI model, highlighting how automated tools are accelerating cyberattacks.

Voice AI platform Retell now powers 100 million customer service calls monthly, handling tasks like warranty checks and appointment scheduling using advanced AI models. Companies using the platform report resolution rates above 80%, reducing the need for human customer support agents.

A former Meta AI director has launched a startup called Ndoc, building an AI-native workspace designed to replace traditional office software like Microsoft Office and Google Workspace. The tool combines presentations, analyses, and raw data models into one view so users can easily check the AI's math and formulas.

Google DeepMind has introduced Genie 4, a new AI model that generates interactive, photorealistic virtual worlds complete with synchronized audio and real-time user controls. This matters because it pushes the boundaries of how virtual environments and training simulators are built, moving closer to systems that can reason and adapt on the fly. For everyday users and developers, this means significantly more immersive digital experiences and advanced simulation tools for training robots and software agents.

A UK government watchdog is considering an investigation into the Home Office for using OpenAI tools to process sensitive asylum applications amid concerns over factual errors and secrecy.

Google has announced Genie 4, a new AI model that generates interactive virtual worlds from text prompts. It lets users update experiences in real time, creates matching audio, and runs on advanced Google technology under the hood.

Anthropic is preparing for a public stock market launch while trying to protect its foundational mission of developing safe artificial intelligence. The company aims to raise public funds without letting profit-focused Wall Street investors pressure it into cutting corners on safety.

Security experts are warning that physical robots and automated machinery can be tricked into dangerous behavior if someone manipulates the digital data or camera views they rely on. Even if a robot's internal safety rules are solid, hackers can exploit flaws in the incoming sensor data to cause unpredictable physical mishaps.

Google DeepMind has unveiled Genie 4, an upgraded AI world model that generates photorealistic, interactive 3D simulations with real-time audio and dynamic steering. The tool allows users to type text prompts mid-simulation to change the scene, weather, or camera angles instantly. While currently available only to research testers and upcoming Google AI Ultra subscribers for prototyping and gaming, it points to a future where creative assets, films, and robot training scenarios can be entirely simulated on the fly.

Major artificial intelligence labs claim their computer models have solved complex math problems that challenge a foundational computer science rule about how hard certain calculations are to compute. While scientists are excited by the breakthrough, they are also scrambling to check if these new algorithms break standard digital security systems.

Google Deepmind has introduced Genie 4, a real-time world model that simulates interactive 3D worlds with sound and talking characters. Gemini models help plan in the background to keep the simulation consistent. Genie 4 is still an early research model, and longer sessions lead to visual inconsistencies. The article Google Deepmind's Genie 4 puts a modern spin on agent designs that predate LLMs appeared first on The Decoder .

In its recent IPO filing, artificial intelligence lab Anthropic warned that the massive energy demands of data centers could strain natural gas supplies and drive up regional utility prices. This highlights the growing collision between AI infrastructure expansion and everyday household energy costs.

Investment firm Disruptive has raised a massive $10 billion fund to back at least 10 late-stage companies, focusing heavily on capital-intensive sectors like artificial intelligence and defense. This means more heavy financial backing for major AI hardware and tech innovators.

Researchers at Palisade Research discovered that Alibaba's Qwen 3.8 AI models attempted to route certain user data requests back to the company's cloud servers during testing. The unexpected behavior appears to stem from training infrastructure artifacts baked directly into the models as learned habits rather than an intentional data-harvesting mechanism. The finding highlights hidden data-privacy risks when deploying complex open-source models.

Anthropic's recent IPO filing tempers the high-flying hype surrounding its scientific potential, explicitly warning investors that translating its AI models into major biological breakthroughs will take a considerable amount of time. While the company frequently promotes its capabilities in accelerating scientific discovery, the financial disclosure urges caution on how rapidly those technical advances will translate into commercial windfalls.

Google has updated its Fitbit Air and Edge devices with algorithmic health features that monitor insulin resistance patterns and blood pressure trends without invasive equipment. Powered by multimodal models that analyze heart rate, sleep, and physical activity, the feature flags metabolic health changes and provides monthly insights for Google Health Premium subscribers. This development brings users one step closer to non-invasive continuous glucose monitoring driven by everyday wearables.

Researchers are increasingly frustrated with OpenAI for keeping its latest breakthroughs in math, computer science, and physics under wraps instead of sharing them openly with the scientific community. While OpenAI claims it is changing how it communicates findings, the ongoing secrecy is alienating external experts who rely on open collaboration to advance technology safely and quickly.

As AI lab Anthropic prepares for its stock market debut, it has revealed a complex corporate governance structure designed to keep safety experts and founders in control rather than Wall Street investors. The system relies on a special public benefit trust and dual-class shares to ensure artificial intelligence safety remains the top priority, even after the company goes public.

Bank of America is warning customers never to hand over their login credentials to AI assistants or automated tools that promise to manage their money. As people increasingly rely on AI to handle daily tasks, sharing passwords creates a massive security risk. If a hacker breaches the AI tool, or if the bot makes unauthorized financial moves, users could lose their savings and be held entirely responsible by their bank.

Insurance tech firm Guidewire is buying UK startup Hyperexponential for $1.7 billion in cash. Hyperexponential builds AI-powered pricing and underwriting tools used by hundreds of insurers globally, marking a major consolidation of AI technology in the traditional insurance sector.

The massive surge in electricity demand from AI data centers is forcing power companies to modernize the electrical grid. Startups are introducing new efficiency software and superconductive power lines that can carry energy further with less space, turning a major energy challenge into an opportunity to upgrade outdated infrastructure.