The ai models & agents stories our desk judged worth your time, written so anyone can follow them.
ElevenLabs now lets human linguists and brand specialists check ads its Ads Engine translates, so small teams can sell in more countries with a safety net.
Read more: ElevenLabs adds human reviewers to its AI tool for translated adsA new Democratic majority in Congress plans to press Nvidia's CEO on China chip sales, which could affect the supply and cost of AI tools.
Read more: Democrats in Congress plan tougher questions for Nvidia on China chip salesResearchers are looking at AI that helps improve itself, and the worry is that people may struggle to check or stop tools that change this fast.
Read more: AI that can improve itself raises new questions about human controlMeta is rolling out updated Muse and Meta AI models that it says forget less and cope better when a task hits a snag. The figures come from Meta's own tests.
Read more: Meta's Muse and Meta AI get better at remembering and finishing tasksVisa ran tests to see how AI agents handle real purchases, and the results could shape whether you let software spend your money.
Read more: Visa put AI shopping assistants to the test with real purchasesA bill from Sen. Bernie Sanders and Rep. Greg Casar would ban super-smart AI and jail some executives, but your chatbot is safe for now.
Read more: Senators propose banning super-smart AI, with up to 20 years in prisonVoters weighed in on 41 data center ballot measures this week and backed tighter limits in all but one. Here is what that could mean for AI and local rules.
Read more: Voters Pushed Back on Data Centers in Nearly Every Local Vote This WeekA test found free AI models that anyone can download come close to a restricted OpenAI model at hacking, at a fraction of the cost.
Read more: Free AI Models Now Rival Locked-Up Ones at Hacking TestsResearchers found more than 800 copies of an AI hacking tool online with no password. The tool has been tied to real attacks on banks and an IT vendor.
Read more: Hundreds of AI hacking tools left open on the web for anyone to useMicrosoft is letting a 2023 deal to rent AI computing from Oracle run out, because the machines use older Nvidia chips. Here is why that matters.
Read more: Microsoft is quietly shutting down AI computing it rented from OracleA YouGov poll for ACI Worldwide finds most UK adults would stop using an AI shopping agent after one mistake, even as tech and payment firms push the tools toward checkout.
Read more: Most Britons would ditch an AI shopping assistant after one errorAndroid lets you swap Gemini for ChatGPT or Claude as your phone assistant. In one test, Claude handled far more phone tasks than ChatGPT.
Read more: Claude can run your Android phone far better than ChatGPT canAnthropic's Claude can now link to your bank accounts and answer questions about your spending. It only reads your data, and it is limited to paid plans.
Read more: Claude can now read your bank accounts and answer money questionsOpenAI's new security AI found 81 high-severity flaws in Firefox. Mozilla has fixed many, so anyone with automatic updates off should update by hand.
Read more: New AI model finds serious security flaws in Firefox before hackers doMeta is pausing new Facebook business terms that could have let it collect linked website content for AI. Page owners should watch for a revised version.
Read more: Meta pauses new terms that let it scrape sites for AIA Pennsylvania program is training state and federal judges to use AI, as courts sort out whether your chatbot conversations can stay private in a lawsuit.
Read more: A Pennsylvania program is teaching judges how to use AI in court workA nonprofit reviewer says Anthropic's AI models now improve about twice as fast, and AI doing its own research is likely the main reason. Here is what that could mean.
Read more: Anthropic's AI is improving twice as fast, and AI is doing the workGoogle's Gemini can now answer plain-language questions from live company databases, using each person's own access. It is in preview for Google Cloud customers.
Read more: Google's Gemini can now answer questions from your company dataGoogle is reportedly preparing Gemini to alert parents when a child talks about self-harm, so chats with the AI may not stay private for kids.
Read more: Google's Gemini may soon tell parents if a child talks about self-harmGoogle is reworking Gemini's sidebar and testing a new overlay animation. Its app may also get an easier way to view and erase your search history.
Read more: Google is changing how Gemini looks and how you clear search historyTaxbit added a built-in AI assistant for tax teams that answers from vetted sources and only within a user's role. Its design offers lessons for any business.
Read more: Crypto tax firm builds an AI helper that cites its sources and respects rolesBusinesses are using more AI, but total spending barely moved in late summer as prices fell. Ramp data shows buyers have the upper hand.
Read more: Companies use more AI but are no longer spending more on itTop AI researchers are leaving OpenAI, Google and Meta to start labs, raising billions on the bet that today's approach alone will not reach AGI.
Read more: Top AI builders are quitting big labs to start their ownAndroid Police argues the Pixel 7 camera still competes in 2026, even though most of Google's newest Gemini AI extras never reached the phone.
Read more: A four-year-old Pixel 7 still takes great photos
