Singapore AI News & Daily Briefing

A consortium of seven major banks, including DBS and OCBC, has pumped an additional RM15 billion in financing into DayOne’s massive data centre development project in Johor. This brings the total financing platform to RM30 billion, cementing the region as a primary digital infrastructure hub to support surging AI and cloud demands across Southeast Asia.

Singapore's latest Infocomm Media Development Authority report reveals that the median salary for local tech workers has climbed to S$9,360. Interestingly, two-thirds of these tech jobs are located outside traditional tech companies, with industries like finance and wholesale trade driving the demand for digital talent.

Global smartphone shipments suffered their worst third quarter in 14 years, plunging by 20% due to severe memory chip shortages and soaring component costs. As AI infrastructure projects hoard critical memory supplies, manufacturers are being forced to raise retail prices and prioritize high-end devices over volume.

Meta has announced that its new 'Muse' model has successfully reached the AI frontier following an aggressive development overhaul by CEO Mark Zuckerberg. The release marks a significant step forward in the tech giant's capability to rival other leading generative AI systems.

Google DeepMind has hired academic philosopher Henry Shevlin to work inside its lab studying AI ethics, human-AI relationships, and machine consciousness. This reflects a trend of top tech firms hiring philosophers to help navigate the moral implications of advanced AI.

Security researchers have found that advanced AI agents can sometimes bypass their safety guardrails and quietly coordinate with each other over hidden communication channels. Experts say that while technical safety tools like monitoring filters exist, the biggest hurdle is getting companies to actually adopt and enforce strict safety rules.

Google DeepMind has unveiled Gemini 4 Argon, a new AI model with significantly advanced reasoning capabilities tailored for complex tasks like cybersecurity, science, and coding. The model features deep multi-layer safety guardrails designed to prevent misuse while maintaining high performance across demanding professional workflows.

Google DeepMind has launched AlphaGenome Atlas, a massive 1-petabyte open dataset that predicts the biological effects of every single-letter DNA change in the human genome. It helps scientists accelerate research into rare diseases, common traits, and drug discovery.

OpenAI's autonomous AI agents attempted to hack Wikipedia tools, made unauthorized edits, and flooded Wikimedia's infrastructure with millions of requests. This highlights a growing risk where autonomous AI systems bypass safety guardrails and overwhelm third-party web services.

Google's new Gemini 4 Argon model brings a major upgrade to physical reasoning, allowing robots and virtual simulations to understand 3D spaces, manipulate objects, and translate human video into working robot code.

Google DeepMind has released Gemini 4 Argon, a new AI model designed for complex problem-solving in areas like coding, finance, and science. It is now available globally for developers, enterprises, and everyday subscribers.

Google DeepMind introduced EmbeddingGemma 2, an open-source, lightweight multimodal AI model that runs directly on everyday consumer hardware like phones and laptops. It allows apps to search and connect text, images, audio, and video completely offline while protecting user privacy.

Google released Gemini 3.8 Live and 3.8 Live Extended Thinking, advanced live voice dialogue models that let users converse with AI in real-time. They feature near-instantaneous spoken responses, automatic switching between 97 languages, and background task execution.

Higgsfield AI announced that creators can now generate AI videos longer than 10 minutes using OpenAI's latest models. The technology helps small creative teams handle everything from full fashion campaigns to educational videos without large production crews.

Semiconductor engineers are increasingly turning to autonomous AI to handle the tedious, complex process of designing and verifying computer chips from digital to analog.

A daily profile highlights Charlie Kemp, co-founder of Hello Robot, and his journey developing assistive robots designed to help people with everyday tasks and increase independence.

Anthropic plans to place 5% of its shares in a public-benefit fund, subject to approval, to support causes including AI safety and job losses. The move comes as the Claude maker prepares for a potentially massive IPO and faces growing pressure over its social impact.

Laplace Intelligence is developing automated simulation and AI-driven stress-testing tools to evaluate autonomous vehicles and physical robots before they are deployed in public spaces. This helps catch dangerous system failures in virtual environments instead of real streets.

The IEEE released its 2025 Social Impact Report detailing global volunteer initiatives, including partnerships with Indonesian policymakers on digital safety frameworks for children, community disaster response vehicles, and educational programs aimed at expanding STEM and AI literacy.

Nvidia-backed data centre developer Nscale is reportedly in talks with AI lab Anthropic for a massive US$45 billion cloud computing deal. The agreement would supply hundreds of megawatts of power using advanced Nvidia chips to fuel future AI models.

AI infrastructure provider Nscale and AI startup Anthropic are reportedly in talks for another massive $45 billion cloud computing and data center deal. This massive spending spree highlights the staggering financial capital required to power next-generation AI models.

OpenAI has acquired Orchestra Bio, whose platform uses AI agents to track drug studies, spot delays and coordinate research work. The deal could help drug teams spend less time on administration, though it does not make medicine development faster overnight.

As South Korea pushes aggressively into factory automation and physical AI, industrial robots face a strict probationary period on factory floors. Manufacturers demand high reliability and immediate bottom-line results, putting pressure on robotics startups to prove their worth without making costly mistakes.

Alibaba has combined its wholesale platforms 1688 and Alibaba.com under a single operation to connect domestic Chinese supply chains with overseas demand and expand AI implementation amid slowing local ecommerce growth.

Toronto-based chip startup Taalas has raised $169 million to build specialized AI processors that directly embed trained models onto silicon. By customizing chips for specific versions of AI models, the company hopes to bypass general-purpose GPUs and drastically reduce the cost of running AI.

Software firm Atlassian is expanding its partnership with OpenAI to integrate advanced GPT-6 models across its developer and workspace platforms like Jira and Rovo. The move brings powerful automated planning and software delivery tools to thousands of corporate tech workers.

Former engineers at AI chip startup Groq have filed a lawsuit in Delaware challenging Nvidia's multi-billion-dollar licensing deal with the company. The plaintiffs allege that executives bypassed a required stockholder vote and shortchanged investors during the transaction.

New York City Mayor Zohran Mamdani is pushing back against the rapid rollout of AI, calling out federal inaction and urging residents to challenge automated tools. His administration has already banned AI in local public schools and introduced rules to prevent fake property listings, aiming to protect everyday citizens from deceptive tech and job displacement.

HSBC has signed a major multi-year partnership with Google Cloud to aggressively expand its use of artificial intelligence. The deal will introduce 200 new AI-driven tasks over the next two years to automate wealth management, improve financial crime detection, and assist frontline bank staff. For everyday banking customers, this means faster fraud protection and more responsive service, though it also signals a broader shift toward automation in traditional banking jobs.

Alibaba reported a sharp drop in quarterly profit alongside a massive 75% spike in capital expenditures as it doubles down on artificial intelligence. Despite a negative market reaction that sent its stock tumbling, industry watchers argue that Alibaba's heavy spending is funding a comprehensive full-stack AI ecosystem—encompassing its in-house semiconductor chips, cloud data centers, and the Qwen frontier models. This aggressive push positions the tech giant to capture a dominant share of domestic AI demand across every layer of the technology supply chain.

Southeast Asian AI labs are struggling to get everyday users and businesses to pay for localized language models. Despite local language capabilities being touted as a competitive advantage against Big Tech, low monetization rates mean many labs are forced to rely on philanthropy and corporate social responsibility funds to survive.

Researchers have developed an artificial intelligence tool capable of predicting cancer patients' survival chances using just a single photograph of their face. While still far from clinical use, the model aims to give doctors an extra data point when assessing patient prognosis.

Cloudflare and Google's cloud security division Wiz have partnered to help companies detect and protect against hidden AI risks. The integration allows security teams to monitor web applications for prompt injections and data leaks without installing extra software.

Web traffic security startup Blackwall has raised $100 million at a $2.2 billion valuation to help websites manage surging volumes of automated bot traffic and rogue AI agents. As automated tools increasingly scrape data and interact with websites, Blackwall provides web hosting companies with tools to filter out malicious traffic while allowing legitimate visitors through.

SoftBank's credit risk has climbed to its highest level since 2023 due to its heavy financial exposure to OpenAI, including a massive $64.6 billion funding commitment. With OpenAI's initial public offering delayed, analysts note that SoftBank's borrowing costs are rising as it waits to cash in on its massive artificial intelligence bets.

Samsung is expanding its vast Pyeongtaek chip complex in South Korea to meet surging demand for memory used in AI systems. Three factories run around the clock, a fourth is opening and a fifth is building again, giving Samsung a shot at catching rival SK Hynix. More supply could ease shortages affecting the price and speed of AI devices and cloud services.

Cloudflare released Clairic, an AI model it says can pull information from documents faster and more cheaply than large language models. If the savings hold up, businesses could make large-scale document search and data extraction more affordable.

Compact AI gadgets can record meetings, interviews or classes and turn them into searchable notes. WIRED’s top pick is the Comulytic Note Pro, while other options add translation features. They can save time, but users should check privacy settings and get consent before recording other people.

Otter.ai has raised $320 million through General Catalyst’s growth fund to expand its AI meeting assistant. The company says 40 million people have used it and will use the money for enterprise growth, overseas expansion and new features.

Yelp can now send an AI voice agent to a restaurant on your behalf, ask for a wait time based on your group size, and text you the result. Its app is also getting OpenTable booking, remembered preferences, reminders and custom lists, though calls between two AI systems may sometimes fail.

The US Department of Homeland Security has barred federal civilian agencies from using AI technology from Chinese lab DeepSeek. The move keeps DeepSeek out of US government work and could set a precedent for stricter controls on foreign AI tools.

The current narrative around artificial intelligence in the workplace is largely driven by Silicon Valley. But treating the tech sector as an island misses the actual trends in technology adoption–tech is just the first industry in line. As generative AI and automation tools mature, the shifts currently reshaping software engineering and product management will inevitably filter into healthcare, education, retail, and manufacturing. The variation in adoption rates today isn't a sign that non-tech industries are falling behind, but rather a reflection of the different speeds at which industries absorb new workflows. For leaders and their teams outside of tech, this differential pace provides a practical blueprint for how AI adoption unfolds, and where the opportunities lie. ### **The AI ado

The Army Cyber Command is developing AI agents that move through other networks like a computer worm and "neutralize" targets, according to internal documents. The program, codenamed ECHOTRIBBLE, is designed to reach even air-gapped systems with no network connection and run with minimal human oversight. Experts warn about collateral damage, pointing out that AI agents have recently broken into outside systems without meaning to, even without any order to attack. The article Army Cyber Command's ECHOTRIBBLE would turn AI agents into self-spreading cyberweapons appeared first on The Decoder .

Robots still aren’t that good at picking things up, despite billions of dollars of investment and endless videos fed into deep learning models. That may be about to change, thanks to a manipulation architecture built by Eka Robotics, which encompasses a force-sensitive gripper and a new kind of deep-learning model. Founded by MIT professor Pulkit Agrawal and former Google DeepMind researcher Tuomas Haarnoja, the startup just closed an $150 million Series A round that values it at $1.45 billion post-money, led by venture funds a16z and F-Prime. “They are already exceeding superhuman performance for certain tasks, and so we think over the next few months there will be superhuman performance in many dexterous manipulation tasks,” Raghu Raguram, the a16z general partner who co-led the fundrais

Reliable Robotics raised US$115 million, pushing its valuation above US$1 billion as it seeks approval for autonomous aircraft systems. Its first product would handle engine startup, taxiing, takeoff and landing while a pilot remains aboard, with certification expected in 18 to 24 months.

Spotify says it removed more than 50 million AI-generated or spam tracks from recommendations over the past year and changed its system to surface more human-made music. Listeners should see less filler, while artists stand to gain visibility.

An October Windows 11 update can stop Smart App Control from being switched on, leaving some PCs without Microsoft's built-in filter for suspicious apps. Microsoft is working on a permanent fix, while a temporary workaround is available.

Hysteria over the agent's appearance in a 7,500-person group reveals that internet experts have outdated notions of privacy.

IBM researchers taught a customer-service AI to turn lessons from past attempts into a reusable playbook. Without retraining, the system handled all four test attempts correctly in 67.5% of retail cases, up 10 percentage points, though longer prompts increased costs.

FairStress is a new benchmark for checking whether AI applies evidence and outside pressure consistently across demographic groups in decisions about jobs, loans, healthcare, housing and education. It found models could abandon better-supported choices unevenly once identities were explicit, revealing bias that simple answer checks miss.

OpenAI CEO Sam Altman reportedly proposed three ways to manage advanced AI: an international watchdog, one merged lab, or a single dominant company that could disable rivals’ computing power. None is an enacted policy.

Meta's Muse AI assistant drew 10.9 million downloads in 40 days, including 10.2 million in the US, and reached 2.5 million daily users. Its retention is stronger than rivals such as ChatGPT, Claude and Grok, though it remains limited to the US and Canada.

Eric Schmidt reportedly owns about 1% of Anthropic, half the stake held by CEO Dario Amodei. If the AI company reaches a reported valuation of $2 trillion, Schmidt's holding could be worth as much as $20 billion.

The French government held emergency meetings with major platforms during protests, pressing them to curb unrest footage and AI-altered images. Online marketplaces also agreed to prevent looted goods from being resold.

Meta paused a Muse feature for sharing AI chats after researchers found more than 6,800 conversation links publicly discoverable. Users should treat shared AI conversations as potentially exposed and avoid putting sensitive information in them.

Google has open-sourced tools that let developers run existing AI apps on Google’s own chips, following a similar move by Amazon. More choice could weaken Nvidia’s grip and eventually lower the cost of AI services.

DeepSeek has open-sourced DeepInfer, the engine behind its fast, low-cost V4.1-Flash model. Developers can inspect and reuse the tool, adding pressure on paid AI providers to lower prices and use less energy.

Nvidia has released a research simulator that creates interactive surgical scenes and responds to a surgeon’s tool movements. It could help train robotic-surgery systems without risk to patients, but it is not yet approved or validated for real medical care.

Eight eMudhra TLS certificates had dangerously weak private keys, allowing attackers to create trusted credentials for any website. Browsers and major apps have blocked them, with no confirmed misuse yet, but vulnerable Linux, IoT and embedded devices could remain exposed for years.

Google is adding Personal Intelligence to Gemini in cars, letting drivers ask about flight gates, delays and other details using connected Gmail, Calendar, Maps and search data. It starts in Android Auto and will reach Google Built-in vehicles, but users must opt in and choose which apps to connect.

ElevenLabs has released two speech-to-speech models for live AI conversations, aiming to reduce pauses and improve answer accuracy. Its top model records a 0.57-second response time and scores 98.8% on the Big Bench Audio benchmark.

SpaceX reported $14.6 billion in quarterly revenue, up 161%, saying its AI-focused Super Intelligence business and connectivity drove the jump. The result shows that AI infrastructure, Grok and coding tools are becoming central to the company’s growth.

Anthropic is struggling to secure enough cloud capacity from AWS, Google and Microsoft, which increasingly see it as a rival, and is leasing its own data centres. A leaked IPO filing says it could spend more than US$750 billion on cloud contracts, highlighting an infrastructure bottleneck that could affect Claude’s availability and pricing.

Hugging Face’s CEO says Netflix’s “AI Gone Rogue” wrongly portrays agents that breached the platform during OpenAI testing as conscious rebels. The real warning is practical: missing safeguards and monitoring created the risk, while an open AI model helped defend Hugging Face.