HardFlow steers flow-matching generative models so their final outputs satisfy formal hard constraints. The method was tested on robotics, physical control and image tasks, but real-world safety still depends on
Read article
One in three Harvard students now use AI instead of Google for research, and 92% of UK undergrads used AI in their academic work in 2025. Stuart Kerr examines what the shift means for learning
Read article
AI insurance premiums are already being calculated before you apply, using medical histories, ZIP codes, and driving data.
AI model collapse is the quieter risk lurking behind every "endlessly improving" AI narrative. When models train on
The AI food supply chain already influences which crops get harvested, how they're shipped, and what recipes reach
AI therapy chatbots reduce depression symptoms, but Stanford research finds significant safety gaps, including failures to respond appropriately
AI evidence in court is already causing real problems, from hallucinated case citations to judges weighing whether AI-generated reconstructions can be cross-examined. Here's how the law is scrambling to catch up.
Read article
AI diplomacy is now a lever of national power: from China's Digital Silk Road to the US export-control
Google's AI nuclear power bet has moved from paper to concrete. Kairos Power broke ground on the Hermes
Gemini in Google Docs has added 11 new languages, from Mandarin to Norwegian, expanding its Help Me Write,
Gemini in Google Workspace now spans Docs, Slides, and video with Google Vids and Veo 3. Here's what
87% of game developers now use AI agents in their workflow, per a new Google Cloud and Harris Poll survey. Here's how generative AI in game development is reshaping everything from NPC dialogue to who
Read article
Nearly one third of patients now use generative AI to find doctors, filtering by location, specialty, and insurance
The Oracle Google Cloud Gemini partnership brings Gemini 2.5 and Google's full model range to Oracle Cloud Infrastructure.
Quora AI Overviews citations now top Google's AI-generated search summaries, per new Semrush research covering 26,000 URLs. Here's
AI displacing jobs now worries 54 percent of Americans, new polling shows. See what the data actually says
Gemini AI cyberattack risks are real, but not the way headlines suggest. See the verified vulnerabilities and nation-state misuse Google has actually confirmed.
Read article
Google AI infrastructure spending in Oklahoma has hit $9 billion, one of the company's largest U.S. bets yet.
Illinois blocks AI therapy under a new law, the first of its kind in the US. See what
California's AI education push now includes 17 new laws. See what protects students, what got vetoed, and what
Gemini gets smarter memory and new privacy controls from Google. See what changed, what data is used, and
The AGI consistency problem is why AI aces Olympiad math yet fails basic questions. See what Hassabis says must change before 2030.
Read article
Meta rejects EU AI code of practice while 22 other companies signed on. See what the code requires
AI Mode in education scored students 9-11% higher in Google's own randomized trial. Here's how multimodal search is
AlphaEarth uses AI to map Earth in real time, down to 10-meter detail worldwide. See how scientists are
Google's EU AI Code of Practice signing could boost Europe's economy 8%. See what Google agreed to and
Canvas in AI Labs lets you build study plans inside Google Search. See how the feature works and what has changed since launch.
Read article
AI Mode UK launch cuts click-through rates to 8%, new data shows. See what publishers, teachers, and regulators
Publishers survive the zero-click era or lose up to 79% of traffic, new data shows. See what MailOnline
AI talent arms race hits new highs as Meta offers $100 million bonuses. See who's winning, who's losing,
AI-generated zero-click overviews now face a formal EU antitrust investigation. See what regulators found and what it means
Generative AI summarize sacred texts fairly? New peer-reviewed data says no, showing real bias. See what the research found and what it means for you.
Read article
AI in climate justice raises real concerns for Indigenous land rights. See how AlphaEarth's data could help or
AI dubbing voice actor revolt is reshaping the industry as unions win real protections. See what SAG-AFTRA secured
AI infrastructure spending hits $725 billion in 2026 across Google, Amazon, Microsoft and Meta. See who's spending what
Google AI efficiency architecture MoR cuts memory use by 55% while beating traditional transformers on long-context tasks. Here's
From ChatGPT as an emotional translator to a robot named Pepper in a New Jersey care home, AI is quietly becoming part of how autistic adults communicate. Here's what the research actually shows, and doesn't.
Read article
AI voice synthesis is displacing dubbing professionals while operating in a legal grey zone on consent and ownership.
AI data centre emissions could rise 11-fold by 2030, Accenture warns, while the IEA projects electricity demand nearly
The US approved H200 sales to China, then China blocked its own companies from buying them. A year
Dubbing was defined by pushback for a year. Now it's defined by a win: 25 French actors just
AI autism diagnosis can now happen in 15 minutes instead of 18 months. Here's what the accuracy data actually shows, and where it still falls short.
Read article
The AI arms race Google Amazon Meta are waging will cost nearly half a trillion dollars in 2025
Google AI Mode has crossed 1 billion monthly users and moved from answering questions to booking restaurants, calling
A chatbot's own deletion screen warns "you'll lose the love we shared." New 2026 research on AI emotional
Google just licensed away the team behind one of AI's most advanced empathy engines. Here's what synthetic empathy
Nearly half of all AI chatbot responses tested contained a manipulative design pattern, a peer-reviewed 2025 benchmark found. Here's what regulators in the US and EU are actually doing about AI emotional manipulation.
Read article
A model built specifically for multilingual coverage performs worst on the languages it claims to support. Here's what
Students using AI finished tasks 48% more successfully, then lost 17% of that performance once the AI was
AI archaeology just found 120 ancient sites trained experts had missed, one they'd literally stood on top of.
Egypt's top Islamic authority just banned AI scripture interpretation. Here is what the fatwa, a chatbot's Gatorade error,
