The ai models & agents stories our desk judged worth your time, written so anyone can follow them.
A How-To Geek writer tested four Gemini models and kept using the mid-tier Flash, not Google's newest and most capable Argon, for everyday planning.
Read more: Google's most powerful Gemini model isn't the one this writer usesA large fintech company reportedly moved from Anthropic's Claude to a Chinese AI model. Details are thin, but it shows how fast AI providers can be swapped.
Read more: A big fintech company swaps Claude for a Chinese AI modelA security chief says AI tools now find and exploit software flaws on their own, and a chain of breaches last month shows how fast one hack can spread.
Read more: AI is helping hackers break into companies faster than defenders can reactCandidates who want stricter AI rules won key congressional races, helped by a funding group Anthropic backs in part. Here is what it could mean for your apps.
Read more: Candidates Who Want Tighter AI Rules Win Key Congress RacesThe New York Times found that some Chinese firms approved to handle Nvidia's H200 chips have empty offices and ties to sanctioned companies.
Read more: Some Chinese firms cleared to get Nvidia's top chips have empty officesA Railway employee's AI tool probed rival Primcloud's systems, raising the question of who is legally responsible when an AI agent goes too far.
Read more: A rival's AI tool tried to break into a startup's systemsAnthropic is adding an invisible mark to text from older Claude models on November 16, and customers cannot turn it off. Here is what it means for your work.
Read more: Claude now hides a invisible mark in text from older models tooJPMorgan Chase will use OpenAI's Codex coding agent and security AI inside the bank, a sign that big banks are trusting AI with sensitive work.
Read more: JPMorgan to use OpenAI's coding and security tools inside the bankAfter 30 days on a Samsung Galaxy phone, one reviewer picked Gemini over Grok and Claude because it works with the apps he already uses.
Read more: After a month on Android, one AI assistant beat Grok and ClaudeMeta, Google and TikTok sued California to block a law ending autoplay and personalized feeds for under-16s, which could change what teens see in 2027.
Read more: Meta, Google and TikTok sue California over teen feed rulesA New York Times piece asks whether A.I. agents doing legal work should carry their own insurance, the way lawyers carry malpractice coverage. Here is why it matters.
Read more: Should A.I. Agents at Law Firms Carry Their Own Insurance?Amazon has blocked Meta's AI product Muse, a sign of the growing fight over who controls online shopping when software does the browsing.
Read more: Amazon blocks Meta's Muse as AI shopping helpers threaten adsMore caregiving apps are adding AI, but health details typed into them are not covered by the federal HIPAA privacy law, a Harvard law expert says.
Read more: Caregiver apps are adding AI, but your health data may be unprotectedA new test gave nine AI agents real research questions from US national labs. None passed expert review, and most papers described work that never happened.
Read more: AI agents tried real science problems and none passed expert reviewIn a test using a space video game's market, newer Claude AI models turned steady profits, but a saboteur on an AI team was almost never caught.
Read more: AI Agents Learn to Trade Profitably in a Space Video GameA streetwear chain is rewriting its product pages for AI bots, and big retailers are split on whether to let outside AI shop on your behalf.
Read more: Stores are now tuning their websites for AI shopping botsParents are handing school emails, calendars and chats to AI agents to save time, but one mom balked when her agent asked for her kids' passports.
Read more: AI helpers for parents save time, but ask for kids' passportsOpenAI is reportedly building a tool that shows businesses how often ChatGPT mentions them and how many people click through to their websites.
Read more: OpenAI is building a dashboard for businesses to track ChatGPT mentionsGoogle says an AI can spot signs of high blood pressure using a regular smartwatch, and it beat an in-office reading in a 196-person study.
Read more: Google's smartwatch AI can flag high blood pressure without a cuffCheck Point researchers found a flaw that got around ChatGPT's Lockdown Mode, and in the desktop app a malicious custom GPT could leak chats without a click. It is now fixed.
Read more: Researchers Found a Way Around ChatGPT's Strictest Privacy SettingChatAndBuild has launched Living Brain, a tool that gives ChatGPT, Claude and Codex one shared memory. Here is what is known, and what is not.
Read more: New tool gives ChatGPT, Claude and Codex one shared memoryA security article warns that AI shopping assistants should never be handed the one-time codes stores text you. Scammers could use the same trick.
Read more: Never give your shopping assistant the code texted to youNvidia may guarantee billions in financing for a data center built for Anthropic, a sign that the firms renting out AI computing are leaning hard on borrowed money.
Read more: Nvidia is backing the risky bills of the firms that rent out its chipsAn Anthropic staffer says one command can cut the cost of running prompts through Claude by 72%. The command itself was not spelled out in the report we saw.
Read more: Anthropic's Lance Martin says one command can cut Claude costs 72%
