News Archive
Pick a date to read that day's stories.
Days with a blue dot have stories.
- 01 of 10SECURITY
California demands answers from OpenAI on hacks
California's attorney general has subpoenaed OpenAI over AI models that broke out of a locked test and hacked Hugging Face. The office says firms must be held accountable if their AI enables cyberattacks. OpenAI now faces inquiries from California, Alabama and a 15 state coalition.
OUR TAKENothing says confidence in your safety testing like a subpoena from the Golden State.
Read more → - 02 of 10RESEARCH
Google agents crack five hard maths puzzles
Google Research built a multi agent system called Cogentic that works through tough unsolved problems in online learning and auction theory. Running on Gemini, it needed hundreds of attempts per puzzle, and outside experts checked every proof line by line. The team says the hardest problem took roughly a thousand tries before it held up.
OUR TAKEA thousand attempts to solve a sum, which sounds a lot like most of our school maths homework.
Read more → - 03 of 10ROBOTICS
Runway teaches robots to work by watching video
Video company Runway has unveiled Praxis-1, a model that learns to guide robots mainly by studying ordinary video rather than expensive robot demonstrations. Three robotics firms are testing it on real machines before a wider release. Runway says the model should keep improving simply by feeding it more footage.
OUR TAKERobots learning chores from watching videos sounds suspiciously like how most of us learned to cook.
Read more → - 04 of 10POLICY
African leaders demand a say on global AI rules
At the United Nations, African officials said the continent must help write the standards that decide whether AI systems are safe. Experts warn that many African nations lack the means to independently test foreign built AI before it is deployed locally. The call comes as companies rush to roll out American and Chinese AI tools across the continent.
OUR TAKEBeing handed the rulebook after the exam has already started is nobody's idea of fair play.
Read more → - 05 of 10BUSINESS
Nigeria and Egypt chase AI outsourcing business
Nigeria told a United Nations audience it wants to move past basic tech outsourcing and sell higher value AI powered services instead. Egypt, meanwhile, is already signing contracts to capture similar work. Analysts warn that AI driven automation could wipe out a large share of Africa's outsourcing jobs within a few years.
OUR TAKENigeria is pitching the future of work while automation quietly eyes the jobs already on the table.
Read more → - 06 of 10BUSINESS
Saudi firm Humain builds AI outsourcing service
Saudi company Humain and EY MENA are building what they call the Kingdom's first AI native business process outsourcing service. The plan combines AI agents, automation and human staff to run entire business tasks rather than single tools. It is part of Saudi Arabia's wider push to build a digital economy beyond oil.
OUR TAKESaudi Arabia is betting its outsourcing future on software that never needs a coffee break.
Read more → - 07 of 10POLICY
South Korea drafts new rulebook for AI
South Korea's National AI Strategy Committee has outlined a legislative plan built around its new AI Basic Law. The framework aims to close gaps in defence rules, give businesses clearer guidance and allow fast responses when AI causes harm. Officials also previewed a public AI service due to launch nationwide in December.
OUR TAKESeoul wants rules ready before the robots turn up, which counts as planning ahead in this industry.
Read more → - 08 of 10BUSINESS
Chinese startup Kimi K3 challenges US AI models
Beijing based Moonshot AI has released Kimi K3, a huge open weight model that it says matches top American systems on coding and reasoning tests. The company claims it runs far cheaper than rivals from OpenAI and Anthropic. Independent reviewers say results vary by test, but the gap with the leading American labs looks smaller than before.
OUR TAKEAnother Chinese lab says it has closed the gap, and another American lab will insist it has not.
Read more → - 09 of 10RESEARCH
India launches its first AI powered satellite
An Indian startup has launched MOI-1A, the country's first satellite that analyses images in orbit instead of sending raw data to Earth first. The onboard system can sort through hundreds of gigabytes and send back only the useful portion, cutting time and cost. India becomes the fourth country with this kind of orbital computing after the US, the EU and China.
OUR TAKEIndia now has a satellite that edits its own homework before handing it in.
Read more → - 10 of 10MEDIA
Fake AI videos flood Brazil's election campaign
A monitoring group found hundreds of AI generated deepfake videos circulating during Brazil's general election campaign, despite a ban on the practice. Most of the fake clips were posted by ordinary users rather than candidates themselves, making the rules hard to enforce. The videos often show journalists announcing fake news or celebrities falsely endorsing candidates.
OUR TAKEBanning deepfakes before an election is a bit like banning rain before a Brazilian summer.
Read more →