The ai models & agents stories our desk judged worth your time, written so anyone can follow them.
Researchers showed a shared ChatGPT link could quietly pull private chats from a signed-in user. OpenAI has fixed it, but the lesson about links still applies.
Read more: A single ChatGPT link could have exposed your private chatsA work step that took 3 days now takes about 90 minutes with Claude, but the AI still stalled on 5 puzzles, so human checks still matter.
Read more: AI cut a 3-day step to 90 minutes, but stalled on 5 puzzlesA researcher says a rogue browser add-on can make Claude's desktop agent run code on your computer. Anthropic has not fixed it, and the weakness reaches other Claude features.
Read more: Anthropic leaves open a flaw that lets a rogue browser add-on control ClaudeScammers are emailing indie studios, demanding money or saying an AI copy of their game will launch first. No working clone has been shown so far.
Read more: Small game makers get threats over AI copies of their gamesAI helpers that shop for you may skip the pages where stores show ads, which could hurt a major source of retail income and change how you shop.
Read more: AI shopping helpers could cut into the ads retailers rely onAgility Robotics is now on the Nasdaq, and its filings reveal what a warehouse humanoid robot costs to rent or buy, which is rarely public.
Read more: Humanoid robot maker Agility goes public, and shows what its robots costGoogle donated three open source tools for running AI agents to a neutral foundation, and says one of them fits 10x more agents on the same hardware.
Read more: Google gives away three tools for running AI agents safely at scaleTwo Windows malware families steal login tokens from AI tools, and those tokens can keep working after a password reset. Here is what to do and who should worry.
Read more: New malware steals AI chat logins that outlast a password changeA group of 200-plus merchants is working with a payments body on rules for AI shopping agents, including which card they use and who is liable when a purchase goes wrong.
Read more: Stores and payment groups team up on rules for AI shopping agentsVisa tested 22 AI shopping agents on everyday purchases. They succeeded only about 57% of the time, so checking their picks before you pay still matters.
Read more: AI shopping assistants bought the wrong thing almost half the timeSalesforce expects AI chatbots to start 20 percent of holiday online shopping visits in 2026, and sellers may need proof a real customer approved each AI-made order.
Read more: AI shopping assistants are about to send you a wave of holiday buyersA Reddit user says AI makes building an app so easy that it can hide the harder job of finding customers, while $200 a month in tools adds up.
Read more: Building an app with AI is easy. Finding customers is the hard partResearchers found a data-theft campaign against South Korean financial firms that leaned on AI coding tools, and the attacker's own files were left exposed.
Read more: Hackers Used AI Coding Tools to Steal Data From South Korean BanksComedian Husk went viral by showing chatbots fail simple tasks. Now OpenAI, Anthropic and xAI want to work with him, which says something about how far AI still has to go.
Read more: A comedian who mocks chatbots now has AI companies chasing himBloomberg explains Meta's Muse and OpenAI's Dots, two AI agents, and what they can do. Know the safety worries before you let one act for you.
Read more: Meta and OpenAI push AI agents, and safety questions followHugging Face's CEO says the biggest AI dangers will come from a few secretive labs, and that people who run AI agents should be legally responsible for what they do.
Read more: Hugging Face CEO says the real AI danger is secret labs, not open modelsMeta says more than 10 million people in the US and Canada set up its Muse AI agent in about six weeks, while AWS sells stores a base for their own shopping helpers.
Read more: Over 10 million people set up Meta's new AI shopping helperOpenAI says it will follow a set of community rules for towns that host its AI data centers, but the details on power bills and jobs are not public yet.
Read more: OpenAI sets out rules for how its AI data centers treat local townsSeenThis says its new tool lets teams brief an AI chatbot and get editable video ads in under an hour, but access is limited and the speed claim is early.
Read more: SeenThis lets your AI chatbot build video ads for youAmazon, Microsoft, Alphabet and Meta plan at least $720 billion in AI spending this year, and investors are asking when it will pay off. Your stock holdings may feel it.
Read more: Big Tech plans $720 billion in AI spending this year, with no payoff yetThe New York Times looks at the experiences that shaped how Anthropic's CEO thinks about AI, a view critics call doomer and others call responsible.
Read more: How Anthropic's chief Dario Amodei came to see AI as both promise and riskAn engineer who helps build Anthropic's Claude Code warns that handing too much thinking to AI can leave you unable to explain your own work.
Read more: Anthropic engineer warns that leaning on AI can dull your thinkingRevolut is reportedly running AI on its own systems instead of paying for Claude, aiming to cut the cost of AI coding help to about a tenth.
Read more: Revolut moves to its own AI to cut coding costs tenfoldA founder known as a great people manager felt lost planning a team of seven staff and over 20 AI agents. Fast Company offers four ways to lead them.
Read more: A startup plans 7 employees and 20 AI agents. How do you manage that?
