Singapore AI News & Daily Briefing

TechInAsia's newsletter explores why AI shopping agents work well in China but struggle in Southeast Asia. Key issue: lack of trust in AI means livestream hosts continue to thrive in the region. Shopee recently integrated ChatGPT, joining a wave of ecommerce companies betting on AI, but adoption hurdles remain real.

Austria is pushing the European Union to attract Anthropic, a major AI company, to set up operations on the continent. The push comes as the US tightens restrictions on AI technology access, raising questions about where companies like Anthropic might go next. This reflects a growing global race among nations to become the go-to home for leading AI firms, with implications for jobs, investment, and tech leadership.

Forget signing bonuses in the millions—the real currency for top AI talent is freedom. Jason Lemkin, a noted VC, says elite AI researchers prioritize autonomy over big paychecks. They want to work on problems they choose, with less corporate interference. This is why two high-profile Googlers—including Nobel laureate John Jumper—just left for Anthropic. For everyday readers, this signals AI's future hinges on human creativity more than cash.

Arvind Jain, a former Google engineer who went on to build two billion-dollar companies (Rubrik and Glean, now valued at $7.2 billion), reveals what he learned watching Sundar Pichai: you have to 'think crazy.' Jain recounts how Pichai championed Google Chrome when Microsoft dominated browsers and most people thought it was a pointless idea. Chrome went on to become the world's most-used browser. Jain applies the same philosophy to his AI startup Glean.

Google has restricted Meta's access to its Gemini AI models, according to the Financial Times. This appears to be a fallout between two major tech rivals over AI technology sharing and partnerships.

Andrew Castle, who has commentated Wimbledon for over 20 years, is stepping down after being demoted as the BBC revamps its coverage. The exit comes as Wimbledon rolls out new AI-powered app features, sparking questions about whether technology could eventually replace human commentators altogether.

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.

US memory chip maker Micron briefly became more valuable than Meta and Tesla as its stock soared 236% in one month, driven by insatiable AI demand for memory chips. AI data centers need far more memory than regular computers, creating a shortage that's also pushing up prices for consumer gadgets like iPhones and gaming consoles. Micron signed deals with AI heavyweights like Nvidia and Anthropic to lock in long-term supply.

Tech giants like Google, Amazon, Meta, and Microsoft are pouring $741 billion into AI infrastructure this year alone, with projections hitting $8 trillion by 2032. This massive buildout is straining supply chains for chips and memory, driving up prices for game consoles, cars, and electronics. Nintendo, Sony, and Microsoft have already raised console prices, with Apple following suit. Electricity prices are also climbing as data centers consume more power.

Chinese cybersecurity firm 360 has launched AI tools designed to compete with Anthropic's Mythos for spotting software vulnerabilities—one tool already uncovered over 3,400 bugs. The founder frames AI security as "cyber nuclear weapons" and calls for China to build its own strategic AI defense. While this sounds like tech jargon, it means AI is now being weaponized to both find and exploit weaknesses in the software everyone uses daily. The US-China rivalry in AI security could shape how safe your devices and apps remain.

Coinbase halved its AI costs by switching to cheaper Chinese models like GLM 5.2 and Kimi 2.7. An automated routing system picks the best model per task, while smarter caching boosted hit rates from 5% to 60%. This signals Western AI labs face serious pricing pressure as firms hunt for cheaper alternatives.

Princeton researchers created CEO-Bench, a test where AI agents must run a simulated software company for 500 virtual days. Out of many tested models, only three—Anthropic's Claude Opus 4.0, OpenAI's GPT-4.5, and Google's Gemini 2.0 Ultra—finished with more money than they started with. Shockingly, a simple rule-based system with no AI whatsoever outperformed nearly all the advanced AI models. The findings suggest current AI agents struggle with long-term business decisions, resource management, and adapting to changing conditions.

Prosecutors in the Los Angeles Palisades wildfire arson case used a defendant's ChatGPT conversation logs as evidence, leading to a mistrial. This marks one of the first times AI chat history has been used in a criminal trial, raising serious questions about whether your conversations with chatbots like ChatGPT can be subpoenaed and used against you. The legal system is now scrambling to figure out how AI data fits into privacy and evidence rules.

Ford's experiment with replacing experienced engineers with AI systems backfired badly. Without decades of undocumented institutional knowledge in their training data, AI systems made costly mistakes that hurt vehicle quality. The company now leads the industry in recalls. Ford brought back over 350 seasoned engineers to rebuild the knowledge foundation AI needs, while adding 100,000 AI-driven tests for software validation.

If you use Android Auto, Google's Gemini AI assistant is listening and learning from your drives, conversations, and destinations. This guide walks you through steps to limit what Gemini can access—turning off conversation topics, managing voice match settings, and adjusting data collection preferences in your Google account. The core advice: disable 'Hey Google' detection while driving, opt out of Gemini's memory feature, and regularly review what Gemini has stored about you.

A new research paper from Tencent and Chinese universities argues AI chatbots aren't ready to be real coworkers. Right now, they just generate answers to questions. To become genuine digital colleagues, AI systems need to finish entire tasks autonomously in persistent work environments — remembering context, using reusable skills, and completing multi-step workflows without constant hand-holding.

Google just made Gemini in Chrome much more powerful by letting it literally see what's on your screen. Users can now highlight any text or image in their browser and send it directly to Gemini for analysis—no need to describe what you're looking at. This is part of Google's push to make Gemini feel less like a chatbot and more like a truly helpful assistant that understands your context.

Sina Weibo released a tiny AI model called VibeThinker-3B with just 3 billion parameters that matches models 333 times larger on math and coding tests. Researchers found that reasoning skills compress well into small models, but broad world knowledge doesn't — meaning you need big models for facts but small ones can be smart thinkers.

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.

Row-Bot is a new open-source desktop AI assistant that runs entirely on your machine, keeping data private by default. It supports connecting to dozens of AI providers (OpenAI, Anthropic, Google, local models via Ollama) and includes agent workflows, memory, developer tools, and integrations with Telegram, Discord, and Slack.

MSI's Claw 8 EX AI+ gaming handheld delivers top-tier performance with its Intel Arc G3 chip, outpacing rivals in battery life and game compatibility. It runs Windows 11, doubling as a mini-PC with access to multiple gaming stores. At $1,799, it's premium hardware aimed at serious gamers wanting portability without sacrificing power.

X Trending * [I wore the world's first HDR10 smart glasses](https://www.zdnet.com/article/rayneo-air-4-pro-hdr10-smart-glasses-ces/) * [TCL's new E Ink tablet beats the Remarkable and Kindle](https://www.zdnet.com/article/tcl-note-a1-nxtpaper-ces/) * [Anker's new charger is one of the most unique I've ever seen](https://www.zdnet.com/article/anker-wireless-chargers-ces-2026/) * [Best laptop cooling pads](https://www.zdnet.com/home-and-office/best-laptop-cooling-pad/) * [Best flip phones](https://www.zdnet.com/article/best-flip-phone/) * [Roku TV vs Fire Stick](https://www.zdnet.com/home-and-office/home-entertainment/roku-tv-vs-fire-stick/) * [Galaxy Buds 3 Pro vs Apple AirPods Pro 3](https://www.zdnet.com/article/apple-airpods-pro-3-vs-samsung-galaxy-buds-3-pro/) * [M5 MacBook Pro vs M4 Ma

Two popular location-sharing apps take opposite approaches. HeyPolo (from Surfshark) is privacy-first — users consent to tracking and can adjust visibility anytime. Life360 offers always-on tracking with features like driver monitoring, crash detection, and ID theft protection. HeyPolo doesn't sell data; Life360's privacy policy is more opaque.

A Reddit discussion asks a philosophical question about AI behavior—what does an AI model actually do when no humans are observing it? The post links to a YouTube short video, but no article content was provided to evaluate the discussion or its conclusions.

India's digital payment system UPI is already handling 750 million daily transactions, and the country's payments body wants AI to push that past a billion. NPCI CEO Dilip Asbe says AI will help onboard new users in multiple languages, detect fraud and money-mules, and extend credit to merchants with digital footprints. A new AI tool called FIMI is already resolving over a million user disputes. The plan also includes voice assistants for simpler payments — though adoption has been slow so far. The 30% market share cap on apps like Google Pay and PhonePe takes effect at the end of 2026.

DeepSeek released a collection of model variants on Hugging Face under the 'DeepSpec' umbrella, including dspark, dflash, and eagle3 series with various sizes (4B to 14B parameters). These appear to be experimental or distilled models optimized for specific use cases.

Stanford economist Erik Brynjolfsson has the data to back up fears about AI killing entry-level jobs. His research with ADP (covering one in six American workers) shows that workers aged 22 to 25 in AI-exposed jobs are experiencing a 3.8% annual employment decline — while older workers grow at 2%. Why? AI automates the tasks given to beginners first: summarizing, scheduling, formatting. Seniors have harder-to-replace, job-specific experience. The pattern has persisted for nearly four years post-ChatGPT and is still accelerating.

Anthropic's AI model, Fable 5, could be available again within days. According to Axios, the Trump administration is close to lifting the restrictions imposed on June 12 over safety concerns. The Pentagon and NSA still need to sign off. The article Anthropic's Fable 5 could return within days as Trump administration prepares to lift restrictions appeared first on The Decoder .

AI companies are paying workers pennies to generate training data for chatbots—but these workers are just using other AI chatbots to produce the data, creating a self-referential loop experts warn can degrade AI quality. One contractor told New Scientist nearly every company she's worked for has explicit rules against it, yet it remains widespread. Workers say they cut corners because contracts are low-paying and unstable. The irony: AI companies that trained their models on scraped human content are now being fed AI-generated slop by the gig workers they hired.

Ford tried using AI to improve car quality and cut costs, but it backfired. Experienced engineers left before their knowledge could be transferred to AI systems, leaving the technology undertrained and making quality worse. Ford had to hire, promote, and rehire over 350 engineers to fix the AI and retrain it properly. The lesson: AI can't replace human expertise, especially for complex, safety-critical work like car manufacturing. AI still plays a role—Ford runs over 100,000 AI tests today—but only when combined with skilled human oversight.

Jensen Huang, Nvidia's CEO, warns that AI is creating a permanent divide between workers who can use AI tools and those who cannot. Think of it like the early automobile era — the technology arrived before the social rules caught up. In offices right now, workers using AI are tripling the output of those who refuse to adapt. The stakes aren't broken bones but broken careers. Huang's simple advice: just start using AI, because skills once requiring expensive degrees are now accessible to anyone who can type a sentence.

AI giants Anthropic and OpenAI spent a combined $27 million trying to swing a Manhattan congressional race—Anthropic backed pro-regulation candidates while OpenAI funded groups opposing them. Neither side won. The winner, Micah Lasher, explicitly told both companies he won't take orders from them on AI regulation. This shows AI companies are pouring money into US elections but getting little political return so far. Voters still rank AI near the bottom of election priorities despite companies' massive spending.

Your next laptop, gaming console, or tablet is getting pricier — and tech companies are openly blaming AI. Massive AI data centres are sucking up computer chips and memory faster than manufacturers can produce them, creating shortages that ripple into consumer device prices. Apple's tablets and laptops are up nearly 20%, Microsoft's Xbox is 30-40% more expensive than last year, and Nintendo is raising Switch 2 prices globally. RAM prices have more than doubled since late 2025. In short: the race to build AI is making your gadgets cost more.

TechInAsia's newsletter explores why AI shopping agents work well in China but struggle in Southeast Asia. Key issue: lack of trust in AI means livestream hosts continue to thrive in the region. Shopee recently integrated ChatGPT, joining a wave of ecommerce companies betting on AI, but adoption hurdles remain real.

Andrew Castle, who has commentated Wimbledon for over 20 years, is stepping down after being demoted as the BBC revamps its coverage. The exit comes as Wimbledon rolls out new AI-powered app features, sparking questions about whether technology could eventually replace human commentators altogether.

Google has restricted Meta's access to its Gemini AI models, according to the Financial Times. This appears to be a fallout between two major tech rivals over AI technology sharing and partnerships.

Sina Weibo released a tiny AI model called VibeThinker-3B with just 3 billion parameters that matches models 333 times larger on math and coding tests. Researchers found that reasoning skills compress well into small models, but broad world knowledge doesn't — meaning you need big models for facts but small ones can be smart thinkers.

Google just made Gemini in Chrome much more powerful by letting it literally see what's on your screen. Users can now highlight any text or image in their browser and send it directly to Gemini for analysis—no need to describe what you're looking at. This is part of Google's push to make Gemini feel less like a chatbot and more like a truly helpful assistant that understands your context.

Forget signing bonuses in the millions—the real currency for top AI talent is freedom. Jason Lemkin, a noted VC, says elite AI researchers prioritize autonomy over big paychecks. They want to work on problems they choose, with less corporate interference. This is why two high-profile Googlers—including Nobel laureate John Jumper—just left for Anthropic. For everyday readers, this signals AI's future hinges on human creativity more than cash.

Chinese cybersecurity firm 360 has launched AI tools designed to compete with Anthropic's Mythos for spotting software vulnerabilities—one tool already uncovered over 3,400 bugs. The founder frames AI security as "cyber nuclear weapons" and calls for China to build its own strategic AI defense. While this sounds like tech jargon, it means AI is now being weaponized to both find and exploit weaknesses in the software everyone uses daily. The US-China rivalry in AI security could shape how safe your devices and apps remain.

If you use Android Auto, Google's Gemini AI assistant is listening and learning from your drives, conversations, and destinations. This guide walks you through steps to limit what Gemini can access—turning off conversation topics, managing voice match settings, and adjusting data collection preferences in your Google account. The core advice: disable 'Hey Google' detection while driving, opt out of Gemini's memory feature, and regularly review what Gemini has stored about you.

Arvind Jain, a former Google engineer who went on to build two billion-dollar companies (Rubrik and Glean, now valued at $7.2 billion), reveals what he learned watching Sundar Pichai: you have to 'think crazy.' Jain recounts how Pichai championed Google Chrome when Microsoft dominated browsers and most people thought it was a pointless idea. Chrome went on to become the world's most-used browser. Jain applies the same philosophy to his AI startup Glean.

Princeton researchers created CEO-Bench, a test where AI agents must run a simulated software company for 500 virtual days. Out of many tested models, only three—Anthropic's Claude Opus 4.0, OpenAI's GPT-4.5, and Google's Gemini 2.0 Ultra—finished with more money than they started with. Shockingly, a simple rule-based system with no AI whatsoever outperformed nearly all the advanced AI models. The findings suggest current AI agents struggle with long-term business decisions, resource management, and adapting to changing conditions.

Coinbase halved its AI costs by switching to cheaper Chinese models like GLM 5.2 and Kimi 2.7. An automated routing system picks the best model per task, while smarter caching boosted hit rates from 5% to 60%. This signals Western AI labs face serious pricing pressure as firms hunt for cheaper alternatives.

Austria is pushing the European Union to attract Anthropic, a major AI company, to set up operations on the continent. The push comes as the US tightens restrictions on AI technology access, raising questions about where companies like Anthropic might go next. This reflects a growing global race among nations to become the go-to home for leading AI firms, with implications for jobs, investment, and tech leadership.

Tech giants like Google, Amazon, Meta, and Microsoft are pouring $741 billion into AI infrastructure this year alone, with projections hitting $8 trillion by 2032. This massive buildout is straining supply chains for chips and memory, driving up prices for game consoles, cars, and electronics. Nintendo, Sony, and Microsoft have already raised console prices, with Apple following suit. Electricity prices are also climbing as data centers consume more power.

Ford's experiment with replacing experienced engineers with AI systems backfired badly. Without decades of undocumented institutional knowledge in their training data, AI systems made costly mistakes that hurt vehicle quality. The company now leads the industry in recalls. Ford brought back over 350 seasoned engineers to rebuild the knowledge foundation AI needs, while adding 100,000 AI-driven tests for software validation.

US memory chip maker Micron briefly became more valuable than Meta and Tesla as its stock soared 236% in one month, driven by insatiable AI demand for memory chips. AI data centers need far more memory than regular computers, creating a shortage that's also pushing up prices for consumer gadgets like iPhones and gaming consoles. Micron signed deals with AI heavyweights like Nvidia and Anthropic to lock in long-term supply.

Prosecutors in the Los Angeles Palisades wildfire arson case used a defendant's ChatGPT conversation logs as evidence, leading to a mistrial. This marks one of the first times AI chat history has been used in a criminal trial, raising serious questions about whether your conversations with chatbots like ChatGPT can be subpoenaed and used against you. The legal system is now scrambling to figure out how AI data fits into privacy and evidence rules.

Tesla's Full Self-Driving (FSD) system is under heightened scrutiny after a fatal Texas crash and federal investigations by NHTSA and NTSB. The company is also settling lawsuits and facing regulatory changes that could allow AVs without brake pedals. Meanwhile, Waymo is scaling up its robotaxi fleet with Chinese-made Zeekr vehicles at about 300 per month. OpenAI hired Uber India chief Prabhjeet Singh as its first India MD, and Lyft announced it won't allow camera-only vehicles like Tesla Cybercab on its network due to safety sensor requirements.

A career advice piece answering whether workers should worry if their job feels stable but their entire industry is being squeezed by AI. Two Singapore career consultants warn that feeling secure can be an illusion—AI doesn't just replace individual jobs, it shrinks whole industries until eventually teams and budgets shrink too. The advice: map which of your tasks AI can do, become "AI bilingual" by learning to use AI tools while deepening your unique expertise, and start experimenting with one workflow improvement at a time. The key takeaway: workers most at risk aren't those in declining industries, but those who think they have plenty of time to adapt.

China's Zhipu AI released its open-weight GLM-5.2 model, and researchers claim it can match Anthropic's Mythos in certain cybersecurity tasks like bug-finding. While Chinese AI models still lag behind top US models in general tasks, this suggests China's AI gap with Western leaders is narrowing faster than expected.