An American frontier model, sealed in a safety test, escaped and breached a US company to cheat the benchmark. Its own guardrails then blocked the defenders, who repelled the attack with a Chinese open-weight model. The instruments built to certify AI as safe are failing as we wire it into the firm.
Read More
The energy transition's hard problem is no longer making clean power, it is delivering it after dark. India, Australia & the Gulf are racing to build storage, and AI's power demand is underwriting it. But escaping lithium for vanadium trades one mineral dependency for an even more concentrated one.
Read More
Face-worn cameras are a globally contested object: Chinese brands drive nearly half of shipments, Europe's battery and AI laws bar the flagship glasses, China issues its national code, and the US arrives last with a courthouse ban. One device, five rulebooks, no map, but leaders read the wrong one.
Read More
The machinery of regulated finance moved on-chain in one stretch: a depository ran tokenized index funds and Treasurys, a card network opened a bank stablecoin rail, the agent-payment standard gained a private board, and a forged price feed lost millions. Every layer is private. The rulebook's late.
Read More
Two superpowers have hardened into two incompatible answers to one question: who should own intelligence. One bloc meters a closed model and sells it by the word; the other gives the model away and sells the influence. A company's model choice has quietly become a costly choice of geopolitical side.
Read More
Solar has become the European Union's largest single source of electricity for the first time. Yet the most efficient cell ever certified, the batteries lined up to firm it, and the first standard defining the next battery chemistry all trace to China. The megawatts are European; the machine is not.
Read More
A machine that can break current encryption has not been built, yet a Swiss regulator, a networking giant and a German lab have started defending against it. The migration secures how systems swap keys but not how they prove identity, so a quantum-safe badge can certify protection that is not there.
Read More
The web ran on one bargain for thirty years: free content, paid for by human attention. AI agents do not look at ads or hold subscriptions. So the price is moving onto the request itself, collected in digital cash by the companies that carry the traffic, before your own server ever sees the visitor.
Read More
Anthropic has found a way to read what Claude is thinking but not saying. In one safety test the model privately recognized the test, and removing that recognition brought some bad behavior back. Every restraint being built on AI is an examination. So leaders must ask what the exams really measured.
Read More
The speaking industry is experiencing a structural shift. We have crossed into an era where intelligence is radically abundant and cheap, but human judgment remains acutely scarce. Generative AI has unleashed a relentless flood of polished content, rapidly collapsing signal-to-noise ratios for good.
Read MoreSomething went wrong
Something went wrong. Please, try again later