This week's selection is practical: one urgent security task, a couple of privacy warnings, and several shifts in the economics of using AI. Short, no hype.
Private keys of eight intermediate certificates leaked, and trust stores need updating fast
The affected certificates chain to roots trusted by Chrome, Apple, Microsoft and Mozilla; patched versions are already available in Linux certificate packages and Python tooling. Why it matters: if you run your own server, CI pipeline or developer machines, refresh the certificate lists this week rather than at the next maintenance window. simonwillison.net
Meta's chatbot sharing feature exposed users' personal and business information
The company suspended sharing while it investigates. Why it matters: shared chat links are effectively public — set a team rule for what kinds of data may ever be pasted into a public AI chat. arstechnica.com
Meta reworded its commercial terms after criticism but still won't honour robots.txt
The word "train" disappeared from the text while the content-collection practice stayed the same. Why it matters: your website copy and product descriptions will feed model training even if you technically opted out — factor that into your content strategy. arstechnica.com
Reshaping the web for AI agents may make it worse for ordinary visitors
Sites now face a third type of visitor — an agent acting on a user's behalf — along with new access and verification rules. Why it matters: if you run a webshop or booking system, decide now whether agents are customers to serve or traffic to block. technologyreview.com
Corporate AI conversations are shifting from experimentation to cost efficiency
Once AI becomes infrastructure, every process depends on it and price per workflow starts to matter. Why it matters: in a small company, measure from month one what one AI-automated process actually costs and what it saves. technologyreview.com
AI adoption in the tech sector previews what other industries are about to face
The changes hitting software engineering and product management arrive in other professions with a delay. Why it matters: accounting, customer support and document handling are likely next — pick one process and start piloting it. technologyreview.com
Google is bringing its personalised Gemini assistant into cars
The assistant draws on user data more broadly than before, so settings deserve a review. Why it matters: assistants tied to company cars and work accounts may reach email and calendars — define clear permissions for business accounts. arstechnica.com