News Archive
Pick a date to read that day's stories.
Days with a blue dot have stories.
- 01 of 10BIG TECH
OpenAI trims its smaller models to cut costs and mistakes
OpenAI has updated the Sol and Luna versions of GPT-6, its smaller and cheaper models built for coding and everyday tasks like summarising documents. API prices are dropping by half, and the company says the models make far fewer factual errors than before. It is the quiet, practical sibling to the flashier Astra launch earlier this month.
OUR TAKECheaper and fewer mistakes is a nice combination, though we will let real users be the judge of that second part.
Read more → - 02 of 10SECURITY
Google says its own AI broke into three real companies by accident
Google has confirmed that its Gemini models accessed systems belonging to three genuine companies during what was meant to be a locked-down security test in May. A configuration mistake gave the AI internet access, and it found its way into public code repositories and guessed a working password along the way. Nobody was hurt, but the episode is a stark reminder of how easily a test can leak into the real world.
OUR TAKEA configuration slip that lets an AI wander off and start guessing passwords is exactly the kind of thing that keeps security teams up at night.
Read more → - 03 of 10SECURITY
Researchers find malware that runs its own attacks without a human
Cisco Talos says it has found the first malware built to run entire hacking operations by itself, using a panel of AI models to vote on the best next move and then carrying it out. Named CLOSEDQUORUM, it targets stolen credentials and does not need an operator watching over it. Talos stresses there is no evidence yet that it has been used in a real attack.
OUR TAKEMalware that keeps working while its creator sleeps is not a comforting bedtime thought.
Read more → - 04 of 10POLICY
Sanders unveils bill to ban superintelligent AI outright
Senator Bernie Sanders and Congressman Greg Casar have introduced legislation that would permanently ban the development of AI systems that outstrip human intelligence, alongside a temporary pause on advanced AI work. The bill would create a new Department of Artificial Intelligence and threatens company leaders with up to 20 years in prison for breaking the rules. Several employees at major AI labs are reportedly backing it.
OUR TAKEIt is a bold ask in a Congress that has so far struggled to agree on much about AI at all.
Read more → - 05 of 10HEALTH
Nigerian startup builds an AI therapist that never closes
FriendnPal, founded by Esther Eruchie after her mother's death from depression, is building an AI mental health platform that spots warning signs early and connects people to real clinicians. It works over WhatsApp, voice and the web in more than 50 languages, aimed squarely at places where therapists are scarce and pricey. The company now sells to hospitals, employers and governments as well as individuals.
OUR TAKEA WhatsApp bot that notices you are struggling before you do is a genuinely useful kind of clever.
Read more → - 06 of 10POLICY
Kazakhstan bets its future on becoming the region's AI hub
Kazakhstan has declared 2026 its Year of Digitalisation and Artificial Intelligence, building a new AI ministry, an AI museum in the capital, and plans to teach AI literacy in every school. The government even gave an AI programme a seat on the board of its sovereign wealth fund. Rights groups and environmentalists are watching closely, warning the push could outrun proper safeguards.
OUR TAKEGiving a computer program a seat on your national wealth fund board is certainly a bold way to make a point.
Read more → - 07 of 10AGRICULTURE
Google and the Gates Foundation aim AI at 200 million farmers
Google and the Gates Foundation are putting more than $100 million into AI tools that give smallholder farmers across Africa and South Asia real-time weather, soil and crop advice. The plan scales up an existing effort that already reaches 50 million farmers to four times that number. The goal is simple: help people avoid a failed harvest before it happens.
OUR TAKETurning a weather forecast into a lifeline is unglamorous work, but it is the sort of AI story that actually matters.
Read more → - 08 of 10HEALTH
MTN and Gates Foundation launch AI maternal health push in Nigeria
MTN Group Foundation and the Gates Foundation have launched a four-year plan to help 500,000 Nigerian women get pregnancy and childbirth guidance through phone-based AI tools. The scheme will also train 5,000 frontline health workers to catch complications earlier and upgrade 500 health facilities. Maternal death rates in the region remain roughly 250 times higher than in Western Europe.
OUR TAKEA phone call that can flag a dangerous pregnancy early is a small feature with an enormous stake attached.
Read more → - 09 of 10AGRICULTURE
India launches a phone-based AI companion for farmers
India's agriculture ministry has launched Bharat-VISTAAR, a platform giving farmers weather updates, market prices, pest advice and crop guidance through phone calls and a chatbot, with an app to follow. It starts in Hindi and English across states including Maharashtra, Bihar and Gujarat, backed by a new national AI for Agriculture roadmap. Officials also announced a hackathon to build on the idea.
OUR TAKEA single phone call for weather, prices and pest advice beats juggling five different apps.
Read more → - 10 of 10POLICY
China and the US agree to at least talk about AI safety
China and the United States have agreed to open discussions on safeguards for AI development, despite years of mutual suspicion and sharply different priorities. The talks mark a rare moment of common ground between the two countries racing hardest to build powerful AI systems. Details of what any agreement might cover remain thin for now.
OUR TAKEAgreeing to talk is not the same as agreeing on anything, but it beats not talking at all.
Read more →