Autonomous AI agents are rapidly becoming the largest attack surface; research shows identity resilience is now the top priority.

A Japanese coastal city accidentally spread a false bear alert with a generative AI-created photo, serving as a wake-up call for stricter verification and protocols around AI in crisis communication.

Anthropic describes how a likely state-sponsored actor used an AI agent to conduct largely autonomous espionage campaigns against dozens of organisations.

Amazon Web Services has announced a new AI service that trains employees in a virtual meeting environment on communication and behaviour.

Baidu open-sources ERNIE 4.5 Thinking, a compact multimodal AI model that must take on GPT 5 and Gemini.

Claude Haiku 4.5 is Anthropic's new compact frontier model that blends speed and scalability with powerful coding and agent capabilities. In this post I explain what the model can do, where it fits, and which practical decisions you need to take as a developer or tech decision-maker.

Google DeepMind is opening an AI research lab in Singapore that will focus on large-scale models, regional applications and AI governance across Asia-Pacific.

Alphabet is investing 40 billion dollars in three new data centres in Texas, Google's largest infrastructure expansion in the US through 2027.

Google is rolling out a new My Stuff hub in the Gemini app, allowing users to manage all their Gemini-generated images, videos and Canvas creations in one place.

Google introduces Nano Banana Pro, a new AI image model within Gemini 3 that offers studio-quality images, improved style control and stricter safety mechanisms.

Google introduces Nano Banana Pro, a new AI image model that promises studio-quality visuals and deeper integration into the Gemini stack.

Google announces Private AI Compute, a hardware-isolated cloud service that lets powerful Gemini models handle privacy-sensitive workloads. The news changes the trade-off between on-device and cloud processing and puts fresh requirements for audits, data governance and vendor architecture on the agenda.

A new strategic partnership between Microsoft, NVIDIA and Anthropic sets the tone for the next phase of generative AI.

A fresh wave of generative AI applications is delivering a striking productivity leap at accountancy firms in the United Kingdom.

OpenAI and Foxconn are joining forces to design and manufacture custom AI datacentre hardware in the United States.

OpenAI and Foxconn are joining forces in the United States to design and manufacture bespoke AI hardware for data centres, a move that further intensifies the race for computing power.

GPT-5.1 delivers higher speed, lower latency and better tooling for developers.

Researchers at Aalto University have unveiled an experimental system that executes complete tensor operations—the backbone of deep-learning models—in a single pass of a beam of light.

Sora 2 marks a leap forward in AI-generated video, allowing creators and businesses to gaze beyond traditional production. In this article I explain exactly what Sora 2 is and share three recent video examples, including the precise YouTube links that demonstrate its possibilities in practice.

One of the most sensitive dossiers in the global AI race is the export of advanced chips to China. Fresh signals from Washington suggest a possible policy shift concerning Nvidia's H200 GPUs, chips that are critical for modern AI infrastructure. This news item summarises the key points and implications.