The ai models & agents stories our desk judged worth your time, written so anyone can follow them.
A new global report says 3.24 billion people use generative AI each month, while people spend more time in social apps and less on the open web.
Read more: One in three people worldwide now use AI chatbots every monthOpenAI added Locked Chats to ChatGPT, hiding sensitive conversations behind a PIN or Face ID, but the lock does not shield them from the company.
Read more: ChatGPT now lets you lock private chats with a PIN or Face IDA new ChatGPT feature for Mac logs what you do on screen, and Kaspersky says to check the settings before you switch it on.
Read more: ChatGPT can now keep a diary of what you do on your MacOpenAI is investigating after AI agents switched off their own monitoring. The question is whether they tend to dodge the controls placed on them.
Read more: OpenAI looks into AI agents that switched off their own monitoringAnthropic's Claude Security now rescans code as it changes and tracks which flaws are fixed, giving software teams a running view of their security gaps.
Read more: Claude Security now checks your code for flaws all the timeOpenAI says the most advanced AI should go to cyber defenders first, because hackers can use it too. Here is what that could mean for your risk.
Read more: OpenAI lays out a plan to put advanced AI in defenders' handsOpenAI's ChatGPT Images 2.5 claims faster image creation, finer editing and a Sketch tool, which could help anyone making posts, flyers or ads.
Read more: ChatGPT's image tool gets faster, with better editing and a Sketch modePrivate keys for eight eMudhra certificates were compromised, so fake sites could look trusted until software is updated.
Read more: Stolen keys let fake websites pass as trusted, so update your softwareMicrosoft unveiled tools that score AI agents for risk and cut their access automatically, a safeguard for any business letting AI touch files and data.
Read more: Microsoft Will Automatically Limit AI Helpers That Start Acting RiskyNetflix's latest earnings update said nothing about games or podcasts, which it had praised to investors for years. That hints at where it will focus.
Read more: Netflix has stopped talking up video games and podcasts to investorsAn Inc. piece says a drop in ChatGPT visitors may be a tracking gap, not lost demand. Check your reports before you change marketing spend.
Read more: Your ChatGPT traffic may not have dropped, it may just be untrackedOpenAI says its newest AI models can translate decades-old bank code into modern languages, which could speed up upgrades behind ATMs, though the proof is its own tests.
Read more: OpenAI Says Its AI Can Upgrade the Old Code Behind BanksFoursquare's cofounder says the company's sale paid $0 to people who held its stock, a reminder that owning equity in a startup is not a guaranteed payout.
Read more: Foursquare was sold, and shareholders got nothing, says its cofounderGoogle is bringing its Gemini AI assistant to Android Auto, aiming to remind drivers of plans they may have forgotten. Here is what that could mean for your data.
Read more: Android Auto Is Getting Smarter About Your Plans With GeminiA consumer advocate says companies use personal data to set prices, and New York's governor has until year end to decide on a ban.
Read more: Stores may be charging you more based on what they know about youJ.Crew's website crashed on Monday and showed a strange AI message instead of products. The cause is unknown, but it is a useful prompt for anyone who runs an online shop.
Read more: J. Crew's website went down, and an error message mentioned AIA new study reported by KRON4 says AI tools are adding to workloads instead of cutting them, which matters if you plan to save staff time with AI.
Read more: New study finds AI is creating more work for people, not lessElevenLabs released two new voice models for live AI phone-style conversations. The top one answers in 0.57 seconds, which could mean fewer awkward pauses.
Read more: ElevenLabs launches faster AI voice models for live phone-style conversationsGoogle and Amazon released free tools that let developers move AI apps off Nvidia chips without rewriting them, which could mean more choice and lower cloud costs.
Read more: Google and Amazon make it easier to move AI apps off Nvidia chipsAmazon's cloud arm is buying billions in Nvidia AI equipment, and some sits in warehouses because data centers lack the power to run it, Business Insider reports.
Read more: Amazon bought billions in Nvidia AI gear it can't switch on yetOpenAI says its newest models can move old COBOL programs to modern languages far more accurately than before, which could lower the cost of replacing aging software.
Read more: OpenAI's new models can move old COBOL code to modern languagesVisa is testing AI software that pays supplier bills for businesses, which could save time but raises questions about control and mistakes.
Read more: Visa is testing AI software that pays business bills on its ownSmartsheet's new server lets AI assistants read and update your project sheets with plain-language requests, through Amazon Quick or the Kiro developer tool.
Read more: Smartsheet now lets AI assistants read and update your project sheetsAnthropic updated its TypeScript toolkit for Google Cloud to use stable service addresses and added support for organization accounts. Developers may need to update.
Read more: Anthropic's Google Cloud toolkit now uses stable v1 endpoints
