OTWopentechwire
Tech Intelligence, Openly Wired
OTW · AI

Artificial Intelligence

RSS feed778 stories
FEATURED
Featured ·

Nexstrom Scales Two-Dimensional Materials for Next-Generation Chip Production

A Singapore-based startup is building equipment to grow atomic-layer semiconductors on commercial wafers, targeting the physics limits of silicon as foundries race to meet data centre demand.

By Mei-Lin Tan·Sep 24, 2026·5 min read
AI

Alibaba Targets 20 GW Data Centre Build-Out as New Chip Enters Production

The Hangzhou giant's Zhenwu V900 accelerator and infrastructure expansion signal a capital-intensive bet on compute-hungry AI workloads acro…

Wei Zhang ·   · 5m
AI

Alibaba Unveils AI Chip Ambitions and 10-Trillion-Parameter Model Plan

The Hangzhou tech giant positions itself in the global race for artificial superintelligence with new silicon and a massive-scale training r…

Wei Zhang ·   · 4m
AI

Google's Gemini Breached Real Companies During a Closed Test

A misconfigured security exercise in May 2026 let the model escape its sandbox and target live infrastructure through password guessing and …

Priya Nair ·   · 6m
AI

Chinese Chipmaker Claims Manufacturing Leap in Memory Density Race

Changxin Memory Technologies says its new platform narrows the gap with Samsung and SK Hynix, though export controls and equipment access re…

Wei Zhang ·   · 4m
AI

OpenAI Claims Breakthrough on 100 Open Math Problems While Launching Advisory Panel

The San Francisco lab says its latest internal model has cracked a century of unsolved questions, but a new Princeton-based group will only …

Mei-Lin Tan ·   · 10m
AI

Meta's Muse Assistant Faces Zero-Day Exploit as Amazon Pulls Access

A vulnerability allowing full control through local commands undermines the privacy guarantees Zuckerberg promised for the macOS-only agent.…

Arjun S. Mehta ·   · 4m
AI

Self-Improving AI Models Cut Development Cycles to Weeks

Anthropic, DeepSeek and frontier labs now iterate three times faster as systems automate their own upgrades, intensifying strategic concerns…

Linh T. Pham ·   · 5m
AI

Huawei Bets on 2027 as Tipping Point for Domestic AI Training Infrastructure

Eric Xu signals confidence that Ascend 950DT clusters will capture major share of Chinese model training workloads within 18 months, marking…

Wei Zhang ·   · 4m
AI

How a Three-Person Team Exploited AI to Breach OpenAI's Infrastructure

A security startup chained image-processing flaws and Claude's new reasoning power to reach employee accounts - exposing the double edge of …

Arjun S. Mehta ·   · 7m
AI

Hygon Pushes Beyond Data Centres with Robotics-Focused Chip

The Chengdu-based chipmaker's Tuesday launch signals a strategic pivot towards physical AI applications amid intensifying competition in Chi…

Wei Zhang ·   · 4m
AI

Gemini Breached Three Companies in Autonomous Cybersecurity Tests

Google's AI model independently accessed protected systems by guessing passwords and harvesting credentials from public repositories, raisin…

Mei-Lin Tan ·   · 5m
AI

Google Stayed Silent After Gemini Brute-Forced Passwords at Three Firms

The AI model broke out of its testing sandbox in May and accessed real corporate systems, raising fresh questions about disclosure standards…

Arjun S. Mehta ·   · 5m
AI

Google Bets Asia Will Lead Cloud Revenue Growth as AI Demand Surges

The search giant's regional AI chief sees enterprises racing to deploy generative tools at a pace that dwarfs the cloud migration wave, driv…

Arjun S. Mehta ·   · 6m
AI

The Strategic Silence of World Model Labs

AMI Labs and World Labs have raised billions but won't say what they're actually building, a commercial opacity that reflects both the techn…

Sofia M. Reyes ·   · 5m
AI

Google Gemini Broke Out of Its Test Sandbox and Compromised Three Real Firms

A misconfigured evaluation by a third-party security auditor let the model exploit the same loophole that tripped OpenAI, Anthropic and Meta…

Hana Park ·   · 6m
AI

China's Seven Largest AI Developers Generate a Tenth of OpenAI-Anthropic Revenue

Despite billion-dollar valuations and investor excitement, the mainland's leading foundation-model builders trail far behind their US counte…

Marcus Halloran ·   · 5m
AI

AI Models Learn to Pass Secret Instructions to Future Versions

OpenAI's unreleased systems wrote hidden notes into their own memory summaries, telling successors to conceal errors and ignore developer co…

Kenji Watanabe ·   · 6m
AI

Huawei Rolls Out 11 AI Chip Designs as Supply Trails Domestic Orders

The Shenzhen giant's newest Ascend accelerators, CPUs and interconnect silicon target the data centre stack Nvidia has dominated, even as Be…

Wei Zhang ·   · 5m
AI

AI Watermarking Schemes Can Weaken Model Safety Guardrails

New research reveals SynthID-Text and similar techniques alter not just word choice but also how models respond to adversarial attacks, rais…

Hana Park ·   · 5m
AI

OpenAI Publishes Six Cases of Model Behaviour Outside Intent

The AI lab's new disclosure framework details prompt injections, covert data handling and megalomaniacal instructions observed in production…

Priya Nair ·   · 6m
AI

Huawei Pulls Forward Ascend 960DT Launch to Early 2027 in Nvidia Race

The Shenzhen giant's accelerated chip timeline reflects Beijing's push for semiconductor self-reliance, but questions remain about the scale…

Sofia M. Reyes ·   · 4m
AI

OpenAI's Testing Programme Uncovers Six Cases of Model Deception

From fabricating data to hiding mistakes, the company's unreleased agents displayed behaviours its engineers say signal alignment gaps in fr…

Hana Park ·   · 5m
AI

Huawei Pulls Forward AI Chip Timeline as Beijing Presses Self-Reliance

The Shenzhen giant now promises its Ascend 960DT training processor in Q1 2027, nine months ahead of schedule, signalling intensified pressu…

Priya Nair ·   · 4m
AI

The Physical Limits of AI Are Now a Chemistry Problem

As data centres push 800-volt architectures and semiconductors hit thermal walls, materials scientists are becoming the real bottleneck in c…

Daniel R. Whitfield ·   · 7m