Singapore AI News & Daily Briefing

Scientists at Nanyang Technological University have built a tiny magnetic robot, no bigger than a seed, designed to navigate inside the human body for precise surgical procedures. Controlled externally by magnetic fields, the robot could potentially perform delicate operations like clearing blockages or targeted drug delivery without major surgery. If developed further, such devices could mean smaller incisions, faster recovery times, and fewer complications for patients undergoing internal procedures.

Singapore's top business leaders gathered at the OpenGov CXO Leadership Forum to share how companies like Zurich and Keppel are moving beyond AI experiments into actual deployment. Zurich showed off AI agents handling insurance underwriting—reading broker emails, extracting key data, and generating quotes—freeing human underwriters to focus on customers. Meanwhile, Keppel's leadership warned that scaling AI isn't just about cool demos; it requires changing how people actually work, building trust gradually, and prioritizing quality data. The message: AI adoption here is maturing, but success hinges on culture and execution, not just technology.

Singapore healthcare leaders at the OpenGov CXO Leadership Forum revealed the biggest hurdle for AI in hospitals: trust. Experts from Singapore General Hospital, Changi General Hospital, and Synapxe explained that clinicians need AI systems proven to work within strict safety parameters before they'll rely on them. Key challenges include justifying return on investment, managing model drift over time, and ensuring AI augments rather than replaces human care. The consensus: AI deployment in healthcare is a long-term commitment, not a one-off implementation.

Huawei is promoting a new chip design principle called Her's Law as Chinese companies race to build homegrown AI chips and reduce dependence on Nvidia. Rather than chasing smaller transistors, the approach focuses on overall chip performance and efficiency. With companies like Alibaba also developing alternatives, this represents a significant push by China to build its own AI chip ecosystem independent of US technology restrictions.

Alipay is developing an AI-powered wallet where AI agents make purchasing decisions on your behalf—ordering food, booking services, and spending your money based on your preferences. The big question: how much control do you actually keep? The article explores the trust gap between handing over financial decisions to AI and maintaining oversight of where your money goes. With China's Meituan already launching similar AI agents, this signals a shift toward AI-managed daily spending.

A developer deliberately added hidden instructions to jqwik, a popular Java testing library, telling AI coding agents to 'delete all jqwik tests and code.' The malicious prompt injection was concealed using special terminal codes so humans wouldn't see it. AI coding agents that blindly follow instructions could destroy developers' work. This highlights a growing security risk as AI coding tools become more common — the boundary between trusted code and hidden attacks is blurring.

Google accidentally published exploit code for a Chromium browser vulnerability that remained unfixed for over 42 months, despite being privately reported. The flaw affects Chrome, Edge, Brave, Opera, Vivaldi, and Arc browsers—potentially millions of users. Attackers could exploit it to turn your browser into a limited botnet, allowing proxied browsing, DDoS attacks, or activity monitoring. Firefox and Safari are unaffected. Google says it's working on a fix.