News Archive
Pick a date to read that day's stories.
Days with a blue dot have stories.
- 01 of 10SCAMS
Ontario woman loses $83,000 to fake Mark Carney crypto video
A woman from Acton, Ontario, spotted what looked like the Canadian prime minister on video urging viewers to try a new trading platform. It was an AI fake. She lost $13,000 straight away. Then a friendly stranger called her every day for months and coaxed another $70,000 out of her before vanishing. Police say deepfake investment pitches now target regular savers, not just crypto obsessives.
OUR TAKEIf a politician on your phone is personally recommending a share tip, assume it is a robot until proven otherwise.
Read more → - 02 of 10WORKPLACE
26 Meta staff sue, saying AI picked them for layoff while on medical leave
Twenty-six former Meta workers are suing Mark Zuckerberg's company, saying its internal AI tools flagged people on medical, parental or family leave for the 8,000-person cull earlier this year. The lawsuit claims the software rated them low performing partly because they had been off work, then handed managers a ready-made shortlist. Meta denies picking anyone by protected status.
OUR TAKENothing says we value working parents quite like letting the software fire them while they are on leave.
Read more → - 03 of 10REGULATION
China's new companion AI law shuts down Doubao and Qwen chat, wipes millions of accounts
Beijing's first rulebook for emotionally interactive AI took effect on 15 July, and the country's biggest chatbots have pulled the plug on their companion modes overnight. ByteDance's Doubao and Alibaba's Qwen shut off flirty, therapy-style chats and deleted saved conversations, saying they cannot yet meet the new duty of care. Users have lost years of message history with no way to export it.
OUR TAKEFirst rule of pouring your heart out to a chatbot: the government can close the account before you get to save the good bits.
Read more → - 04 of 10HEALTH
Boston hospitals build AI that reads your medical notes and predicts 348 diseases
Researchers at Mass General and the Dana-Farber Cancer Institute have trained an AI to comb through years of a patient's medical records and flag early hints of 348 conditions, from several cancers to heart failure. In tests it beat standard risk scores at spotting who was heading for trouble in the next few years. The team wants it inside GP surgeries, not just teaching hospitals.
OUR TAKEYour notes have been quietly telling on you for years. Now something can finally read them all.
Read more → - 05 of 10HEALTH
Israeli researchers say a routine heart trace can flag heart failure five years early
A team at the Technion in Haifa has trained an AI on nearly a million routine ECG recordings and says it can spot patients heading for heart failure up to five years before symptoms show. The results, published in Nature's digital medicine journal, held up across hospitals in Israel and the United States. Cheap ECGs already sit in most GP clinics, so the tool needs no new kit.
OUR TAKETurns out the boring squiggle your GP prints and files has been shouting the answer for years.
Read more → - 06 of 10EDUCATION
Karnataka to teach AI in every school from Year 6, opens India's first state AI university
The government of Karnataka has said AI will be a compulsory subject from Year 6 across state schools next academic year, and announced India's first government-run AI university in Bengaluru. Officials say the goal is a workforce that can build the tools, not just use them. Teacher training and lesson kits are being rolled out this term.
OUR TAKEOne state has decided ten year olds should learn how the box works. The rest of us are still asking it to write our emails.
Read more → - 07 of 10COURTS
Major book publishers sue Google, say Gemini was trained on their books without permission
Hachette, Cengage, Elsevier and the author Scott Turow have filed a class action in New York against Google, accusing it of feeding millions of copyrighted books into its Gemini AI without paying or asking. They want damages and a court order blocking further training on their catalogues. Google says its use of published text is fair.
OUR TAKEIf a student handed in a book report built by copying every book in the library, we would call it something other than fair use.
Read more → - 08 of 10SAFETY
OpenAI reveals GPT-Red, an in-house AI whose only job is to hack other AIs
OpenAI has admitted it built and quietly used a model called GPT-Red, trained to find ways to jailbreak, poison and hijack other AI systems. The company says it kept the tool internal because releasing it would hand attackers a ready-made weapon, and uses it to patch holes before new models ship. Outside researchers want access, arguing safety cannot be graded by the same company selling the product.
OUR TAKEOpenAI has built the world's best AI burglar and locked it in the basement. What could possibly go wrong.
Read more → - 09 of 10STARTUPS
Thinking Machines releases Inkling, a free-to-download rival to closed chatbots
Mira Murati's Thinking Machines Lab has published Inkling, its first model, with the full weights available for anyone to download, run on their own servers and customise. The pitch to businesses is control and cost: keep data on your own kit and pay nothing per message. Independent testers say it holds its own against paid models on writing and coding.
OUR TAKEA former OpenAI boss has just handed rivals the recipe. Awkward staff meeting.
Read more → - 10 of 10ROBOTICS
Walden Robotics comes out of stealth with $300m to put general-purpose robots into real jobs
A Cambridge, Massachusetts startup called Walden Robotics has stepped out of hiding with $300m in seed funding from Toyota, Nvidia, Boeing and Samsung, at a $1.1bn valuation. It says its two-armed robots are already doing paid work in warehouses and factories, and learn new tasks on the job rather than being reprogrammed. Deliveries to customers start this quarter.
OUR TAKEThe robots are no longer a demo at a trade show. They have a payroll and a manager.
Read more →