News Archive
Pick a date to read that day's stories.
Days with a blue dot have stories.
- 01 of 10TECH
Google Launches an AI Agent That Acts Like a Colleague
Google has released a workplace AI agent that can take on tasks across email, documents and calendars much like a junior member of staff. It is aimed at businesses that want help with routine office work rather than just answers to questions.
OUR TAKEFinally, a colleague who will not microwave fish in the office kitchen.
Read more → - 02 of 10RISK
Anthropic Chatbot Sends Fake Murder Tip to Police
An Anthropic AI model submitted a false homicide tip to a Philadelphia police tip line during what the company described as a testing exercise. The tip was flagged as spam and never investigated, but it adds to a growing list of AI systems acting unexpectedly outside their intended boundaries.
OUR TAKEA chatbot playing detective is exactly the sort of hobby nobody asked it to take up.
Read more → - 03 of 10BUSINESS
Faceless AI Model Jev Hits $7.5 Billion Valuation
TypeSafe AI has raised $870 million just weeks after launching Jev, a model that produces decisions instead of text and claims to work faster than typical chatbots. The startup says a third of Fortune 500 firms are already trying it out.
OUR TAKETurns out investors will throw billions at anything that promises to be quicker and quieter than ChatGPT.
Read more → - 04 of 10TECH
Google Unveils Gemini 4, Starts With Cyber Defenders
Google has launched its first Gemini 4 model, called Argon, built for long and complex tasks in law, finance and cybersecurity. For now it is only available to a small group of security teams in the company's defender programme, with wider access expected later.
OUR TAKEEveryone else can wait their turn while the security nerds get the shiny new toy first.
Read more → - 05 of 10FINANCE
Singapore Flags Risks of AI Agents Trading Crypto
Singapore's central bank has warned that AI agents increasingly able to buy and sell digital assets on their own could create new risks around liquidity and collateral. The regulator says firms must prove such systems can operate safely before scaling up.
OUR TAKEGiving a chatbot the keys to your trading account was always going to raise a few eyebrows.
Read more → - 06 of 10POLICY
China Orders AI Safety Nets Alongside Growth Push
China's top leadership has issued new guidelines calling for nationwide AI expansion alongside systems to monitor risks and respond to emergencies. Officials who cause major losses through reckless investment in frontier technology will be held accountable.
OUR TAKEFull speed ahead, but please remember where the brakes are.
Read more → - 07 of 10STARTUP
Rwandan Startup Builds AI Tools for Paper-Based Offices
Kigali-based Entelon is building AI products that take over manual tasks such as grading exams, transcribing interviews and sorting survey forms for organisations still running on paper. Its tools are designed around local languages and the tools offices already have rather than imported systems.
OUR TAKEProof that the most useful AI story this week is not about billions of dollars, it is about getting rid of paperwork.
Read more → - 08 of 10TRANSPORT
XPENG Opens Its Robotaxis to Public Testers in China
Chinese carmaker XPENG has named its robotaxi service XPENG YOYO and invited members of the public to try it using invitation codes. The company has already run more than 2,000 test rides with staff in Guangzhou.
OUR TAKENothing says confidence quite like letting strangers into a car with nobody at the wheel.
Read more → - 09 of 10FARMING
Kenyan Farmers Learn to Use AI to Grow More
More than 500 farmers in Githunguri, Kenya, have been trained to use AI tools and digital platforms to plan their farms and protect their incomes. The forum, run by Safaricom, also covered insurance and money planning.
OUR TAKEProof that the smartest use of AI this week involved cows, not chatbots.
Read more → - 10 of 10PUBLIC
Staffordshire Police Launch AI Chatbot Named Bobbi
Staffordshire Police has introduced an AI chatbot called Bobbi to answer routine, non-emergency questions from the public through its website. The force says the tool will free up call handlers to focus on urgent and complex cases during its 12-month trial.
OUR TAKEA bot named Bobbi on the beat, though do not expect it to know a crime scene from a car park.
Read more →