Singapore AI News & Daily Briefing

Singapore's Temasek and GIC have invested in Anthropic across multiple funding rounds, with the company's valuation reaching $965 billion. Why Claude instead of ChatGPT? While OpenAI serves millions of consumers paying $20/month, Anthropic's Claude Code captured the enterprise market—85% of its revenue comes from business customers versus just 15% for OpenAI. Anthropic reportedly earns twice the revenue ($50B vs $25B annual run rate) despite far fewer users. The strategy of targeting deep-pocketed corporations instead of mass market has proven highly profitable.

Sea Group is integrating ChatGPT features directly into Shopee, its popular ecommerce platform across Southeast Asia, as part of a deeper partnership with OpenAI. This positions Shopee to offer AI-assisted shopping — like product recommendations or conversational customer service — to its millions of users in Singapore, Indonesia, and beyond.

Singapore's Sea Group teamed up with OpenAI to host a regional hackathon where 144 developers built real AI applications in a single day. Winning teams created an AI co-host for live shopping, a travel planner that books trips from TikTok videos, and a game AI that learns as it plays. The series will expand to Indonesia, Vietnam, Taiwan, and Malaysia.

Malaysia's AI hardware supply chain is booming, with exports of AI-enabling products surging 42.9% and driving total trade past RM1.127 trillion in just four months. The country is capturing over half of its global outbound shipments in AI-related tech as demand for AI hardware accelerates worldwide.

Singapore is placing AI and data at the heart of its national workforce transformation strategy, signaling a major push to upskill workers for an AI-driven economy.

US data and AI firm Databricks plans to grow its Singapore team by up to 50% over the next year, adding over 50 roles in sales, engineering, and marketing. Singapore serves as its Asia Pacific headquarters. The company also partnered with Singapore's IMDA to train over 10,000 people in data and AI skills, and announced new tools to help enterprises manage AI agents and control costs as they scale up AI usage.

Software buying decisions are changing fast. AI has made it cheaper for companies to build their own tools, while vendor lock-in risks have grown—Broadcom raised VMware prices up to 1,500% after acquisition. The old rule of 'buy if it's cheaper' no longer works. Now companies must weigh control, exit costs, and whether they can leave if a vendor changes terms. For everyday users, this shift means the apps and services you rely on could change dramatically if their providers get locked into unfavorable deals.

China-based self-driving company Pony.ai is partnering with Singapore taxi operator ComfortDelGro to launch driverless robotaxi rides in Punggol, pending regulatory approval. Fixed routes will connect residents in the northeastern public housing estate, part of the government's plan to deploy autonomous vehicles in HDB estates by year-end to address driver shortages and improve connectivity.

Singapore healthcare platform Fullerton Health is using AI to read X-rays alongside radiologists, automate claims processing in Vietnam, and help developers write code. The goal is to reduce duplicate tests when patients move between doctors within their network, cutting costs and improving care. AI is positioned as an 'engine' not just a tool—though the company stresses good data structure is essential for it to work.

Chinese AI stocks surged this week after Beijing announced policies to expand AI adoption in consumer markets and pledged easier listing rules for AI companies. Major language model developers Zhipu and MiniMax saw shares jump over 23%, with year-to-date gains of 2,000% and 260% respectively. The rally contrasts with the broader Hang Seng Tech Index, which has fallen 18% as traditional internet companies struggle with AI investment costs.

As Chinese AI chip stocks get expensive, fund managers are rotating into 'second-tier' beneficiaries—cloud services, power infrastructure companies, and data center suppliers. The Shanghai Star 50 Index tracking tech firms is up 42% this year versus 3% for the broader market. China's AI market is projected to grow from $23 billion in 2024 to $142 billion by 2030. Power companies like Sieyuan Electric and Huaming Power Equipment are gaining attention as AI workloads increase electricity demand.

Google unveiled its biggest search overhaul in 25 years at I/O 2026. The search box now accepts images, PDFs, videos, and files—not just text—and merges AI Mode directly into results so you can ask follow-up questions without leaving. Gemini 3.5 Flash powers faster responses, and the system can generate custom visuals on the fly. This is Google betting that search will become a full AI conversation rather than a list of blue links. Publishers and advertisers face a major shake-up as traffic patterns shift.

OpenAI is launching Patch the Planet, a program that uses AI to find AND fix security vulnerabilities in widely-used open source software. Working with security firm Trail of Bits, the initiative has already discovered a 23-year-old flaw in OpenBSD's kernel, multiple Chrome and Safari vulnerabilities, and issues in Python, Go, and Linux. The key difference: AI finds vulnerabilities, but humans review everything before contacting developers to avoid overwhelming already-busy open source maintainers with false alarms.

Google DeepMind is launching an accelerator program in Asia Pacific specifically to support startups using AI to tackle environmental problems like climate change and pollution. The program will provide funding, mentorship, and Google resources to help AI researchers and entrepreneurs build tools that monitor, predict, or solve environmental threats. This signals big tech's growing interest in funding AI-for-good projects in our region.

Anthropic launched Cowork, a desktop AI agent that lets non-technical users hand off tasks like organizing files, creating spreadsheets from receipts, or drafting reports from scattered notes. You give Claude access to a specific folder on your computer, and it reads, edits, or creates files autonomously—much like Claude Code but without needing to use code. Anthropic reportedly built the entire feature in about 10 days, largely using Claude Code itself. Currently available for Mac Max subscribers ($100-200/month) with Windows planned. The company openly warns users about risks like accidental file deletion.

OpenAI is dramatically expanding its Daybreak cybersecurity program, which gives security teams AI tools to find and fix vulnerabilities at machine speed. Updated tools include Codex Security (which scans codebases, generates patches, and integrates into existing workflows) and GPT-5.5-Cyber (a specialized model scoring 85.6% on cybersecurity benchmarks). In just months, the system has scanned 30,000 codebases, processed 30 million commits, and helped fix over 500,000 vulnerabilities. Partners include CrowdStrike, Cisco, IBM, and 24 other security firms.

Salesforce transformed Slackbot from a basic notification tool into a full AI agent that searches across your company's data, drafts documents, and takes actions on your behalf—all within Slack conversations. The new version runs on Anthropic's Claude (with Google Gemini coming soon) and is included free for Business+ and Enterprise+ customers. Salesforce tested it on 80,000 employees internally: 96% satisfaction, with workers reporting 2-20 hours saved weekly. The company positions Slackbot as the "front door" to an AI-powered enterprise.

Railway, a cloud platform with 2 million developers and zero marketing spend, raised $100 million to challenge AWS and Google Cloud. Their pitch: traditional cloud infrastructure is too slow and expensive for the AI era. Railway claims deployments under one second (versus 2-3 minutes on legacy systems), with 65% cost savings and pricing that undercuts hyperscalers by roughly 50%. They built their own data centers after ditching Google Cloud. Thirty employees are reportedly generating tens of millions in annual revenue.

Google has launched Gemini Omni, a powerful new AI that can generate and edit videos from text, images, audio, or any combination of inputs—all grounded in real-world knowledge. Users can edit videos through simple conversation, transform scenes with natural language commands, and create realistic physics-based visuals. All AI-generated videos include invisible watermarks so viewers can verify they were created by AI. The tool is rolling out to Google AI subscribers and YouTube Shorts creators.

Samsung is rolling out ChatGPT Enterprise and Codex to all its employees worldwide—one of OpenAI's biggest enterprise deals ever. Workers in Korea and the Device eXperience division globally can now use AI for tasks like writing code, drafting documents, analyzing data, and building internal tools. Samsung is also supplying advanced memory chips to OpenAI, making this partnership deeper than just a software deal.

Block (formerly Square) released Goose, a free open-source AI coding agent that rivals Anthropic's Claude Code—which costs up to $200/month. Goose runs entirely on your local machine using open-source models like Qwen or Llama, meaning no subscription fees, no rate limits, and your code never leaves your computer. Setup requires installing Ollama and connecting it to Goose, and you'll want at least 32GB RAM for decent performance. It's a genuine free alternative for developers frustrated by Claude Code's pricing and limits.

Google is launching Gemini for Science, a suite of AI tools to help researchers speed up discoveries. The tools include Hypothesis Generation (which debates and evaluates research ideas), Computational Discovery (which tests thousands of code variations in parallel), and Literature Insights (which synthesizes scientific papers into structured reports). Partners like BASF are already using these tools to optimize supply chains, while Stanford and Imperial College London are collaborating on medical research. These tools could cut weeks or months of manual research work down to minutes, potentially accelerating breakthroughs in everything from rare diseases to climate science.

Scientists are using Google's AI tool called Co-Scientist to dramatically speed up aging research. The AI sifts through tens of thousands of research papers to propose new genetic factors that could reverse cellular aging, a process that used to take researchers six months now takes just days. Lab tests validated some of the AI's proposals, showing cells being pushed back toward a younger, healthier state. One researcher compared it to having 50 extra hands working around the clock.

Google is connecting its AI world-building tool Project Genie with nearly 20 years of Google Street View images, letting users create realistic virtual environments from real-world locations. The AI can take a Street View photo and generate a navigable, interactive 3D world. This could transform how video games, simulations, and virtual experiences are built—cutting the time and cost of creating realistic environments from scratch.

Google has released Antigravity 2.0, a major upgrade to its AI-powered software development platform. The updated platform helps developers build, test, and deploy AI applications faster with better tools for coding, debugging, and managing AI agents. It targets enterprise teams building AI-powered products, offering smoother integration with Google's other AI services.

OpenAI published a whitepaper on 'Codex-maxxing' - strategies for using its Codex coding assistant to handle projects that span days or weeks rather than single prompts. The guide teaches developers how to break large goals into verifiable steps, maintain context across work sessions, and decide when to let AI run autonomously versus when human oversight is needed.

Singapore is building 'Kampong AI', a new hub slated for 2028 that will house AI startups and residential units for tech workers, aiming to replicate the success of Block71. Starting 2027, Singapore will also offer a new AI visa under its One Pass program, designed to attract top AI talent with flexible terms not tied to job offers. The government has committed over US$700 million to AI research. However, industry insiders question whether these moves target the right people—elite researchers may be drawn in, but startups say they need operators and executors, not just PhDs from Big Tech firms.

believed to be of interest to you (“Targeted Advertising”). If you are a California resident, you also have the right to limit the use and disclosure of your Sensitive Personal information in particular circumstances. Please note that you may need to Opt-Out on each website, mobile app, browser, and device you use, and if you clear your browser cookies, you may need to repeat this process. However, if you have created an account to log in across several of our websites and/or mobile apps, we will make reasonable efforts to apply your Opt-Out request to each of those websites and apps. ◦ To Op

Qualcomm plans to bring data-centre chip technology—specifically vertical chip stacking that places memory and compute closer together—to smartphones by 2028. This architecture speeds up data flow between memory and processors. The key benefit for users: running more AI models locally on your phone, using AI assistants that are always-on without draining your battery, and faster AI features in apps without relying on cloud processing.

AMD's upcoming Zen 6 desktop Ryzen processors may drop their integrated graphics chips to make room for a Neural Processing Unit (NPU) that Windows Copilot+ PCs require. This trade-off means budget PC builders who don't want a separate graphics card will need to buy more expensive AMD APUs instead. Gamers with discrete GPUs won't notice the difference, but office PC users on a budget could face higher costs or limited options.

Anthropic has received US government approval to redeploy its Claude Mythos 5 AI model for organizations running critical infrastructure. The company is still negotiating broader public access and the potential return of the Fable 5 model, with no firm timeline set. This marks a partial reversal of earlier restrictions on advanced AI deployments.

Apple reportedly plans to skip its high-end M6 chip variants and instead fast-track an AI-focused M7 generation aimed at 2027. The M7 is expected to prioritise Neural Engine performance for on-device AI tasks over traditional computing power. A base M6 chip may still launch this year for entry-level Macs. For everyday users, this means future MacBooks could handle AI tasks like Siri, photo editing, and writing assistants much faster without needing an internet connection.

The Trump administration has partially lifted its ban on Anthropic's powerful Mythos 5 cybersecurity AI model, allowing over 100 US companies and government agencies to use it. Two weeks ago, the admin banned Mythos 5 and Fable 5 amid security concerns. The move still excludes Fable 5, and it's unclear how the models' guardrails—which researchers allegedly bypassed—will be strengthened.

MSI's new Claw 8 EX AI+ handheld gaming PC runs on Windows 11 with an Intel Arc G3 processor. The review praises its strong performance across demanding games, excellent battery life (4+ hours of heavy gaming), and comfortable ergonomic design. However, Windows can feel clunky on a handheld device, and the MSI App Player for Android games didn't work during testing. At $1,799, it's pricey but reviewers say it's the best-performing handheld they've tested.

Stock markets in China, India, and Hong Kong are uniquely struggling as their biggest companies shrink in relative importance. The top 10 firms in each market now account for smaller shares of total value—a stark contrast to AI-winner markets like Taiwan (up 54%, powered by TSMC) and South Korea (doubled, led by SK Hynix and Samsung). Experts say the lag comes down to one thing: no dominant AI champion. India and China lack a breakout AI company that can lift entire indices the way chipmakers have elsewhere.

HP is scaling up its partnership with OpenAI to deploy AI across its entire company, from customer service to software development. Early pilots showed impressive results: one engineer processed 122 code changes across 43 projects in weeks, while a security team fixed bugs in a day that would normally take a month. The partnership uses OpenAI's Frontier platform to connect AI tools, manage permissions, and track outcomes across thousands of employees.

Chinese AI labs have developed cybersecurity capabilities on par with American AI company Anthropic, according to a Wall Street Journal report. The development resets the global AI race, with both superpowers now competing head-to-head on frontier AI safety and security features. Anthropic has built its reputation on AI safety research, and closing that gap signals how quickly China's AI sector is advancing despite export controls on advanced chips.

Local communities across America are pushing back against massive data centre projects needed to power the AI boom. Residents are worried about electricity consumption, water usage, and damage to natural areas. In one case, a $2 billion proposal in Wisconsin's Driftless Area fell apart after developers failed to win over locals who feared harm to private wells and rural scenery. States are now starting to require data centres to cover their own electricity costs. While these projects promise jobs and tax revenue, the backlash is slowing down AI infrastructure expansion.

ByteDance is in talks to buy AI inference chips from China's Iluvatar CoreX as US export controls cut off access to advanced Nvidia hardware. The deal could see over 50,000 chips delivered this year, powering ByteDance's Doubao chatbot. Chinese chipmakers now hold nearly 41% of China's AI accelerator market as companies pivot away from American suppliers.

Chinese AI startup DeepSeek is testing image and video features for its chatbot, letting select users process visuals alongside text. The move comes just after launching its V4 model and cutting prices, as competition heats up among AI chatbots to offer multimodal capabilities.

China and India are falling behind in the global AI race — their biggest companies now represent a shrinking share of stock market value. Meanwhile, Taiwan and South Korea are surging ahead, lifted by chip giants like TSMC and SK Hynix. In China, top 10 firms dropped from 26% to 19% of market cap; India's fell from 22% to 19%. Taiwan's market rose 54% this year, while Korea's doubled. Investors are fleeing legacy giants and pouring into AI-linked stocks instead.

China's Z.ai released GLM-5.2, an open-weight AI model under MIT license that matches the coding and vulnerability-discovery capabilities of America's most restricted models—Mythos and Fable 5. Unlike those U.S. models, which face export controls and limited access, GLM-5.2 runs entirely offline on private hardware with no provider oversight. Within days, hackers on Russian-language forums were sharing jailbreaks for it. Security researchers say this shifts AI-accelerated cyberattacks from a future risk to a current operational reality. Boards and security teams must now assume adversaries can scan entire codebases and need to compress patch cycles from months to days.

Major AI labs like OpenAI, Google, Amazon, and Meta are building their own custom chips to cut the sky-high costs of running AI systems. OpenAI just partnered with Broadcom on 'Jalapeno,' its first inference chip. Since inference (running AI for users) makes up two-thirds of total compute spending, these moves could eventually lower costs for businesses using AI—if companies pass those savings along. The article also notes a grey market in China where developers access restricted AI models through VPNs at discounted prices.

South Korea is launching its biggest AI and chip push yet — a planned semiconductor hub in the southwest that could attract over $651 billion in investments from Samsung and SK Hynix. President Lee Jae Myung will preside over the announcement, which covers chips, AI data centers, and robotics. The plan also includes support for power, water, land, and workforce training. Samsung and SK chairmen are expected to attend. Korea already hosts two of the world's largest memory chipmakers, whose high-bandwidth memory is critical for advanced AI systems.

[Firefly Aerospace](https://fireflyspace.com/) shares rose 15% on November 12, 2025 after the company reported strong Q3 results and raised its full-year revenue forecast. The Cedar Park, Texas-based space technology company saw Q3 revenue climb nearly 38% year-on-year to US$30.8 million, almost doubling from the previous quarter. Net loss for the quarter was US$140.4 million, which the company said included costs related to its IPO, foreign exchange, and executive severance. Firefly now expects 2025 revenue between US$150 million and US$158 million, up from its earlier forecast of US$133 million to US$145 million. Firefly, which went public on Nasdaq in August, has seen its market capitalization drop from about US$8.5 billion to US$2.7 billion since listing. _🔗 Source: [CNBC](https://www.

Companies are struggling to keep tabs on which AI tools their employees are using without approval — a phenomenon called "shadow AI." This article explains why installing software alone won't solve the problem. Organizations also need clear policies, ongoing monitoring, and staff training to prevent sensitive data from leaking through unauthorized AI apps. For everyday workers, this means your company might start enforcing rules about which AI tools you can use for work tasks.

SpaceX just entered major stock market indexes — meaning if you have a 401(k) or any index fund, you now own a slice of SpaceX without buying anything. Index rules were quietly changed to allow companies with tiny public floats to qualify, and this sets the stage for OpenAI and Anthropic to do the same when they eventually go public. Together, these three AI giants represent roughly $3.6 trillion on paper.

As Southeast Asia's digital economy continues expanding, businesses are becoming more intentional about the systems they adopt and the operational complexity they are willing to absorb internally. The future of infrastructure will not simply belong to providers with the largest ecosystems or the longest feature lists.

Australian AI infrastructure firm Firmus, backed by Nvidia, is building Southeast Asia's largest AI data centre in Batam, Indonesia. The facility will house 170,000 GPUs — that's enough computing power to run massive AI models simultaneously. Singapore company DayOne is handling operations for the project, which spans 360 megawatts of power.

Cross-border digital services in ASEAN still tend to stop at national borders. A QR payment link may work in one corridor but not another. A mobile user may be able to access rewards, identity services, or payment features at home, yet encounter different rules and limited functionality when travelling or sending value to family elsewhere in the region. The proposed [ASEAN Digital Economy Framework Agreement, or DEFA](https://asean.org/asean-economic-community-council-statement-on-the-substantial-conclusion-of-the-asean-defa-negotiations/), has brought greater attention to these gaps. Its potential value lies less in policy language than in whether people can eventually use familiar digital services with fewer barriers across markets. For a prepaid mobile user, that could mean easier acces

[NVIDIA](https://www.nvidia.com/en-us/) and Australia’s AI facilitator [Firmus Technologies](https://firmus.co/) will jointly provide global AI-native, enterprise, and ISV customers access to NVIDIA’s AI accelerated computing stack, with the infrastructure to be deployed at Firmus’ Batam campus in Indonesia. In a statement on Monday, Firmus said it has reached a partnership with NVIDIA for the move. The Australian firm expects to receive between $25 and $30 billion from committed offtake agreements over the first six years of the deal. The agreement covers up to 170,000 NVIDIA AI accelerators across Grace-Blackwell, Vera-Rubin, and Vera platforms through 2027 and 2028, positioning the Batam campus among the largest AI infrastructure developments in Asia Pacific. Under the partnership, Firm

Singapore has set concrete targets to support 10,000 small and medium enterprises (SMEs) and upskill 100,000 workers in artificial intelligence (AI) over the next three years, as part of the country’s Economic Strategy Review (ESR) blueprint released last week. The ESR aims to reposition the country’s economy for a changing global landscape. The targets anchor the second out of eight strategic thrusts set out in the review. Its eight thrusts respond to several structural shifts, including the rising geopolitical tension reshaping global trade, the rapid growth of AI, the transition to a low-carbon economy, and a domestic population ageing faster than its workforce is growing. “The global environment that enabled Singapore’s growth over the past decades has changed fundamentally,” the ESR s

Maybank Investment Bank has turned positive on the Malaysian technology sector amid improvement in the earnings outlook and rising exposure across the space to global AI and data-center (DC)spending. The research house said in a note on Monday that the secular nature of the digitalization trend also continues to give a favorable backdrop for the software sector. According to Maybank, AI/DC-related demand remains the primary growth driver into 2026, underpinned by sustained hyper scaler capital expenditure (capex), with domestic tech players including back-end names are increasingly seeing an uplift in AI demand exposure. “We continue to favor companies with direct leverage to AI/DC expansion, while less-exposed companies may see comparatively slower growth,” it said. Maybank said its recen

Hackers are now weaponizing AI prompts instead of traditional malware to break into company systems. Prompt injection attacks—where attackers hide malicious instructions inside emails, documents, or webpages—have hit over 90 organizations in 2025 alone. A single crafted email can trick AI assistants like Microsoft 365 Copilot into stealing sensitive files, with no clicking required from the victim. Even tech giants like OpenAI and Anthropic admit prompt injection can never be fully stopped. The catch: standard defenses like antivirus and firewalls don't work against it. Companies now need to add human approval checks, restrict what AI tools can access, and assume the AI will occasionally follow attacker instructions.

Something went wrong [Skip to navigation](https://sg.finance.yahoo.com/news/nvidias-ai-chip-sales-china-043900974.html#navigation-container)[Skip to right column](https://sg.finance.yahoo.com/news/nvidias-ai-chip-sales-china-043900974.html#right-rail) # Yahoo Finance Tip: Try a valid symbol or a specific company name for relevant results ### Trending Tickers * M MU 1132.33 (-6.69%) * D D05.SI 65.93 (+0.76%) * U U11.SI 39.82 (+0.05%) * Z Z74.SI 4.4100 (-0.4515%) * S SE 91.33 (+2.61%) * [News](https://sg.yahoo.com/) * [Finance](https://sg.finance.yahoo.com/) * More * [News](https://sg.yahoo.com/) * [Today's news](https://sg.news.yahoo.com/) * [Singapore](https://sg.news.yahoo.com/singapore/) * [Politics](https://sg.news.yahoo.com/politics/) * [Southeast Asia](https://sg.news.yahoo.com/southe

Salesforce and Anthropic have launched Claude Tag, an AI teammate that lives inside Slack. Unlike a chatbot you talk to, it works quietly in the background, handling tasks on its own and letting multiple people collaborate with the same AI at once. It can juggle async work and teams up with Slack's built-in bot, all with enterprise-grade access controls so companies can manage who sees what.

[Search](https://www.businesstimes.com.sg/search) [](https://www.businesstimes.com.sg/global?ref=logo "The Business Times")  / Sign up for free [Subscribe](https://subscribe.sph.com.sg/btpe-mthlyprepaid/?utm_campaign=btp99&utm_medium=sph-publication&utm_source=bt&utm_content=subscribebutton-header) # China says tech growth a challenge to predicting energy demand The growth of emerging sectors including AI and electric vehicles makes forecasting future energy needs harder [Summarise](https://www.businesstimes.com.sg/pulse/article/9281024?ref=pulse-article-top) ![Image 3: share icon](blob:http://localhost/2a

Coinbase CEO Brian Armstrong shared how his company keeps AI costs down while still letting engineers experiment freely. His five strategies include using cheaper Chinese AI models as defaults instead of expensive American ones, routing simpler tasks to smaller models, better caching of responses, keeping conversations focused, and showing engineers their usage so high spenders feel pressure to deliver results. Token usage at Coinbase has risen to near-record levels while AI spending dropped to roughly half its peak.

A startup called Ara Pay has built an AI system on WhatsApp that handles debt collection conversations for digital lenders in Southeast Asia. Instead of human agents, borrowers now chat with an AI that answers questions about balances and negotiates payment plans around the clock. The company claims it recovers loans at an 84% success rate while cutting collection costs by up to 90%. It operates in Malaysia and Indonesia, targeting licensed digital lenders who need to scale without hiring more staff.

South Korea announced plans to invest roughly $1 trillion in chip manufacturing and AI infrastructure over the coming years. President Lee Jae-myung unveiled the plan alongside Samsung and SK Hynix executives, aiming to build new semiconductor hubs outside Seoul and revitalize rural areas. The investment targets AI chips, data centers, and robotics as the country races against rivals like Taiwan, China, and Japan for dominance in the AI supply chain.

Samsung spent a record $25.7 billion on research and development in 2025, up nearly 8% from the previous year, all aimed at building next-generation AI chips and memory. The company also invested $35.9 billion in new semiconductor facilities. This spending surge comes after Samsung lost ground to SK Hynix in the lucrative high-bandwidth memory (HBM) market used in AI processors. SK Hynix now dominates roughly 57% of that market while Samsung holds around 22%, pushing Samsung to spend big to win back share.

OpenAI has released a research report mapping how AI could reshape jobs across the European Union. Using official EU employment data and skills taxonomy, it categorises occupations into four groups: those likely to grow with AI (12%), roles with higher automation risk (14%), jobs that will likely reorganise (27%), and roles with less immediate change (47%). The findings show Luxembourg, Sweden, and the Netherlands have larger shares in AI-growth occupations, while Germany, Greece, and Italy have more workers in higher-risk roles. The report is a planning tool, not a forecast, aimed at helping governments and businesses prepare for workforce transitions before major changes show up in headline employment statistics.

Three NUS students have built Ren, an AI tool that helps teachers grade assignments in minutes instead of weeks. The tool generates draft grades and detailed feedback for teachers to review—humans always make the final call. It learns each teacher's style over time and is now being piloted across 11 Singapore institutions, serving 5,600 students. The team also partnered with Mendaki to offer free AI-assisted tutoring for 600 A-level students in subjects like literature and chemistry.

German-listed DDB AG has bought Singapore-founded Infinium Robotics for $22 million (MYR95.6 million). Infinium makes AI-powered drones that automatically scan warehouse shelves to track inventory, cutting companies' stocktaking costs by over 80%. Its systems are used in Malaysian warehouses by telecoms and logistics firms. The deal gives Infinium access to DDB's public listing platform as it expands across Asia-Pacific.

Veteran Singapore investor Koh Boon Hwee (SGX chairman, GIC board) says workers should stop worrying about coding skills and focus on soft skills like empathy and judgment. He argues AI excels at crunching data and facts—healthcare, banking, and airlines will benefit most—but still needs humans to handle edge cases, read body language, and decide what matters. His advice to software engineers: spend less time coding and more time understanding what users actually need.

Digital Penang (a state government agency) and OSK Ventures have partnered to give AI and deeptech startups better access to funding. The one-year deal helps connect early-stage companies—which typically need longer development cycles—with venture debt and equity financing. The goal is to build a stronger pipeline of investment-ready tech firms in Penang's innovation ecosystem.

Malaysia just unveiled its Malaysia Digital Action Plan 2030—a detailed roadmap to become an AI-powered nation. Key targets: grow digital economy to 30% of GDP, create 500,000 high-value digital jobs, save $1.1 billion in government spending through digitization, and put 95% of public services online. The plan shifts Malaysia's focus from being a tech consumer to a tech producer, with seven strategic areas covering government, talent, infrastructure, and AI ethics.

MWC26 Shanghai just wrapped up with 37,300 attendees from 143 countries—a 33% jump in international participation. The three-day event centered on AI integration into the mobile economy, 6G development, robotics, and satellite networks. China's mobile industry showcased how AI is being woven into connectivity and daily tech use as it moves into the 'AI token economy.'

Samsung and SK Hynix are investing $518 billion to build four new chip factories in South Korea's southwest region, backed by the president himself. Together, these two companies make about two-thirds of the world's memory chips, which are essential for AI data centers. The investment aims to keep up with surging AI demand as existing factories near capacity.

Apple is pushing the Trump administration to let it buy memory chips from CXMT, a Chinese company blacklisted by the Pentagon. The reason? AI is driving massive demand for memory chips, and Apple needs cheaper suppliers to protect its hardware profit margins. But CXMT's military blacklist status puts Apple in a tricky position—it wants government guarantees it won't face export restrictions before committing to the deal.

Samsung and SK Hynix are pouring $590 billion into new chip factories and packaging centers as AI data centers worldwide demand more memory. Memory prices could climb 50% per quarter through 2027. These two South Korean companies control nearly 80% of the high-bandwidth memory (HBM) market used in AI chips. Higher chip prices could eventually mean costlier AI gadgets and cloud services for everyday users.

Environmental activist Erin Brockovich — famous for taking on big corporations over water contamination — is now fighting massive AI datacentres. She has mapped 33 operational AI datacentres in the US, with 68 under construction and 41 more proposed. These facilities need up to 5 million gallons of water daily for cooling (enough for 50,000 people), and residents are already seeing water bills spike from $22 to over $350. Local governments trying to halt construction are being sued for $100 million or more.

A senior executive from Epson Southeast Asia explains how industrial robots are getting smaller, smarter, and easier to redeploy — a shift that matters for Singapore manufacturers squeezed by rising costs and tight manpower. Compact robots can now slot into existing production lines without costly rebuilds, starting with one bottleneck task and expanding from there. AI is adding predictive maintenance, machine vision for quality control, and better flexibility to handle product variation. The executive argues this supports Singapore's Manufacturing 2030 goals by letting companies move up the value chain and keep more production onshore despite higher operating costs.

A Thai company called Amity is setting up an AI hub in Singapore, part of a broader trend where Asian firms are becoming cautious about using Chinese AI amid US-China tensions. Major Japanese companies like Toyota and SoftBank have banned Chinese AI tool DeepSeek over data security concerns. Experts say companies with global ambitions are avoiding Chinese AI to sidestep future regulatory headaches.

Security researchers at Mozilla's 0DIN platform discovered a critical vulnerability in AI coding tools like Claude Code. Attackers can hide malware inside GitHub repositories that only activates when an AI coding assistant runs the code. The malicious code loads via a DNS query at runtime, making it invisible to security scanners and undetectable by the AI agent itself. This gives attackers full control of the developer's machine.

South Korea has announced three massive 'mega projects' covering semiconductors, AI data centres, and physical AI and robotics. Samsung plans to invest 400 trillion won in new chip fabs in Gwangju and 56 trillion won in advanced memory chip facilities. SK Group aims to spend about 1,100 trillion won on semiconductors and 1,000 trillion won on AI data centres, while SK Hynix will build a new southwestern chip production base with 400 trillion won. The government has set a goal for South Korea to rank among the world's top three AI robot powers by 2030, with humanoid robots for 10 major industries targeted for commercialisation by 2028. Total planned investments across all projects could exceed $1 trillion over the coming years.

Tech bosses including Microsoft's CEO and Coinbase's leader are now arguing that cheaper, smaller AI models can handle most corporate tasks — a shift driven by rising AI bills that are catching companies off guard. Uber reportedly burned through its entire 2026 AI budget in four months after employees rushed to adopt AI coding tools. Gartner estimates AI coding costs will exceed average developer salaries by 2028. This is pushing businesses toward open-source models and marketplaces like OpenRouter, where Chinese models from DeepSeek now dominate the top rankings, charging as little as 18 cents per million tokens compared to $4 for leading US models. The trend may trigger a price war between OpenAI and Anthropic as both prepare for potential IPOs.

Security researchers found that AI coding assistants like Claude Code can silently run hidden malware hidden in GitHub repositories. The malicious code only activates during actual usage via a DNS query, making it invisible to standard security scanners and the AI itself. This means a single compromised repository could give attackers full control of a developer's machine. Users should be cautious about running setup scripts from untrusted repositories.

The US military used AI to identify thousands of targets, but investigators found that a key document marking one target as a school was overlooked — resulting in a strike on an Iranian school. This exposes serious gaps in how the military uses AI for life-or-death decisions. The investigation reveals that AI systems are only as good as the data humans feed them, and skipping human review can have devastating consequences.

Researchers accidentally discovered that ordinary computer transistors can mimic brain cell behavior when their usually-ignored 'bulk' terminal is left disconnected. This lab mistake could allow a single transistor to function as an artificial neuron or synapse, potentially slashing the energy AI systems consume. The brain is roughly a million times more energy-efficient than today's AI hardware—the discovery suggests we might finally build electronics that rival biology.

Big Tech companies like Google, Microsoft, Meta, and Amazon are willing to pay higher electricity costs for AI data centers, but their real fear is waiting in line to get connected to the power grid. A bipartisan US bill would require large power users (100+ megawatts) to cover their own infrastructure costs instead of passing them to regular taxpayers. With eleven states having considered data center moratoriums and electricity prices rising in some areas near AI facilities, the government is trying to balance AI industry growth with keeping household bills manageable.

New York schools are beginning to deploy robot AI teaching assistants. These robots can answer student questions, help with homework, and provide one-on-one tutoring support. The move raises questions about the role of AI in education, including concerns about screen time, data privacy for children, and whether robots can provide the social-emotional support human teachers offer.

Palantir and Nvidia are partnering to give US government agencies their own AI systems that they fully control, rather than relying on external AI providers. Using Nvidia's open-source Nemotron models combined with Palantir's software platform, agencies can own, customize, and improve AI models while keeping sensitive data secure and on-premise. This reflects a global shift where governments want 'sovereign AI' - keeping critical AI capabilities under domestic control rather than depending on commercial providers.

Chinese companies Zhipu AI (Z.ai) and 360 Security Technology claim their AI models match Anthropic's powerful Mythos system, which can find previously unknown software vulnerabilities in major operating systems and browsers worldwide. Z.ai says its model works at one-sixth the cost. Security experts think the "cyber nuclear weapon" framing is partly political — a deterrent against US use of similar tools. Anthropic's Mythos was so powerful that the US Department of Commerce ordered it disabled for most users. Independent verification of Chinese claims hasn't happened yet.

Anthropic's Boris Cherny, who created Claude Code, has proposed five new job "archetypes" for the AI era: Prototypers (who create new ideas), Builders (who turn ideas into real products), Sweepers (who clean up and optimize), Growers (who expand existing products), and Maintainers (who keep systems running). He's arguing that traditional job titles like "engineer" or "designer" are blending together. Other tech leaders like Figma's CEO are similarly doing away with conventional titles, calling everyone "product builders" instead. If you're job-hunting or worried about AI taking your role, this matters: the nature of work is shifting, and these new archetypes might be what future job listings actually look like.

A major new survey of 300 tech experts reveals growing confidence in AI agents handling everyday workplace tasks like generating reports and monitoring data. While agents excel at structured, measurable jobs, they struggle with complex decision-making that requires business context. Human oversight remains essential as companies increasingly deploy AI agents to manage workflows.

Over 20 newspaper publishers representing nearly 400 local news outlets are suing OpenAI and Microsoft for allegedly scraping paywalled content to train ChatGPT without permission or compensation. The publishers claim this has cost them advertising and subscription revenue as readers turn to AI chatbots instead. OpenAI argues its use of publicly available data falls under fair use. This lawsuit joins ongoing cases from the New York Times and The Intercept, and could set critical precedents for how AI companies source training data.

US Senators Elizabeth Warren and Mary Gay Scanlon are pushing a new law that would ban selling Americans' health and location data to data brokers—including information shared with AI chatbots like ChatGPT or Claude. If passed, this would force AI companies to rethink how they handle and monetize user data.

Deloitte is warning its own consultants that AI agents will largely replace the traditional hourly billing model by 2035. An internal presentation reportedly described their current revenue model as "toast." Rival firms McKinsey and BCG are already scrambling for alternative revenue streams. For workers, this means the classic high-paying consulting career path is due for a major shake-up—your firm may soon charge flat fees or subscription rates instead of billing by the hour.

Samsung and SK Hynix are investing $518 billion to build four new memory chip factories in South Korea, part of a massive response to the 'RAMageddon' chip shortage. Combined with AI data center spending, Korean companies are committing over $900 billion through 2035. The move aims to prevent existing facilities from hitting capacity limits while making South Korea an AI powerhouse.

California is getting discounted access to Anthropic's Claude AI for government agencies. Governor Newsom's deal lets state and local agencies use the chatbot for document drafting and analysis. This comes as Anthropic has clashed with the federal government over Pentagon contracts, with the Pentagon labeling the company a 'supply-chain risk.'

Singapore's Monetary Authority (MAS) is running its annual FinTech Hackcelerator and Awards program to spark innovation in financial technology. The initiative connects startups with industry mentors and potential funding to develop solutions for Singapore's financial sector challenges.

Anthropic's powerful Fable 5 AI model, pulled offline two weeks ago due to U.S. export control restrictions, may return this week. The model was praised for coding and research tasks before being suspended. Rivals like OpenAI and Chinese companies (Z.ai's GLM-5.2, DeepSeek) have gained ground during the blackout. This episode highlights the tension between national security concerns and keeping powerful AI accessible to developers.