Pope Leo XIV called on October 2 for the Church to renew its alliance with artists and cultural institutions, arguing that the spread of artificial intelligence makes it urgent to distinguish human art from machine-generated production. Speaking to participants in a meeting organized by the Vatican’s Dicastery for Culture and Education, he described a difference that precedes judgments of beauty, contrasting art with statistical calculation based on images made by others. He also warned about cultural homogenization. His intervention gives artists institutional support in a dispute often framed around speed and cost. It expresses a philosophical position without establishing a ban on artistic uses of AI.
For institutions that commission art, this position raises practical questions about their next contracts. A parish choosing an image for a new chapel can buy a generated illustration or pay an artist to develop a work with the congregation. That choice affects who earns a living and who answers for the finished image. The Pope’s call leaves the practical terms of the renewed alliance unspecified. Artists can use digital tools while still negotiating payment, attribution and control over how their work is reused. If Church patrons follow his appeal, their purchasing decisions will need to give human authors a place in the budget. A commissioned artist should be able to see that commitment in the contract and the agreed fee.
On October 2, Apple said it will add controls to macOS’s Full Disk Access, a permission originally intended to let backup apps work around other privacy protections. The company warned that some apps use it in ways that could expose files, email, messages and browsing history without users fully understanding the reach. For communication apps, Apple said, the risk can extend to the privacy of people on the other end. It promised that granting this level of access will require a “very explicit” action, but gave no timetable or design details. Apple linked the change to increasingly capable and autonomous AI agents.
One account holder’s approval can affect another person. A message sits on two devices, but the recipient chooses which apps may inspect it on their Mac. An AI agent with broad access could draw on private records well beyond a task such as organizing files; Apple’s announcement, however, does not accuse any named product of misusing the permission. The practical test for the new controls will be whether the request explains what an app can reach and lets a person grant only the access needed for a task. Until Apple shows the design, users cannot tell whether the planned action will clarify that boundary or simply add another confirmation to a broad permission.
On September 29, Meta expanded Muse for small businesses, adding connections to professional Instagram and Facebook accounts and tools such as QuickBooks, Shopify and Stripe. The company says owners can ask the agent to review sales, flag expenses, draft campaigns and sort incoming tasks. It also says nothing is published, sent or paid for without the owner’s approval. Meta’s vice president of AI products told Axios that about a third of Muse’s early users had already connected some kind of business account. That is an early adoption signal, though Meta has not published its own active-user figures or how often connected accounts lead to completed work.
The new connectors bring the agent into a shop’s working records, from sales and customer questions to advertising performance and bookkeeping. For an owner without separate marketing or finance staff, one interface spanning those chores could free time for running the business. It also puts sensitive records and routine decisions within reach of a service run by the same company that sells advertising and controls platforms where many shops find customers. Meta says the owner approves actions affecting the outside world; each business still has to decide what data Muse can access and which changes it may make. The reported connections show owners granting access. Meta has not published figures on hours saved or sales generated by Muse.
On September 30, the US Centers for Disease Control and Prevention released its review of 39 models that forecast weekly influenza hospital admissions during the 2025–26 season. Google’s SAI-FluEns ranked first among individual team submissions. The public forecast the CDC used was a separate median ensemble, which ranked seventh overall. At the national peak, weekly admissions exceeded 40,000; fewer than 25 percent of that ensemble’s two-week prediction intervals across jurisdictions contained the observed figures. Flu forecasts can help officials plan hospital beds, staffing and antiviral supplies, making the distinction between a contest winner and a public health forecast tangible.
The report shows accuracy declined when flu activity changed quickly. A season-wide score can hide how a model performs in a state or around a sudden surge, just when a hospital may need to revise staffing or bed plans. The review is a retrospective comparison; it does not say Google’s winning model has replaced the CDC’s ensemble or that hospitals used it to make those decisions. The next test is whether forecasts communicate uncertainty clearly and update local estimates early enough to inform response before admissions climb.
