In 2025, we put together the Guidelines for the Responsible Use of Generative AI in Aged Care through the AI Adoption in Aged Care Workgroup, led by the Multicultural Communities Council of SA. They're a plain-English, Tier 1 document — a short guide for staff using everyday tools like ChatGPT, Claude or Gemini, built around four principles. More than 120 organisations have adopted them since.

The 2026 update (available since last week) adds a new operational layer — a quick-reference set of ground rules covering which tools are approved, what should never go into a public AI tool, when AI can draft and when a person has to check before anything goes out, and who to ask when someone's unsure.

They're free for any provider to download and adapt. If you do one thing with them, run a short workshop with your team around the template — that's where informal use starts becoming managed use.

Author Lily Dunn's mother survived a head-on crash, and never once reached for the word "trauma". Meanwhile Dunn's university students were emailing to say a bad day had left them too traumatised for class. Her essay is about a word stretched so wide it's losing its meaning. Read it here.

In aged care the word travels the other way. The people reaching for it are often the younger workers. The people with the strongest claim to it — older, often of diverse backgrounds, frequently having gone through armed conflict, displacement or a hard migration — are the ones who use it the least. Dunn worries the word is wearing thin from overuse, but in our corner it often goes missing from the people it was meant for.

The trap is reading that silence as an answer, because it misleads in both directions. Take it as someone coping quietly and you can miss a person who isn't. Take it as buried trauma and you can saddle someone who's managing perfectly well with a wound they don't have. You can't tell which from the quiet. Working out which one is in front of you is most of the actual job, and nothing about the setup makes it easier — the system listens for distress in one register, the clinical one, and plenty of people never spoke that language. They show it through the body, through family, through faith, through getting on with it.

The hardest version is the distress that never gets counted at all. A sudden crisis can move a couple who've spent a whole life together into separate places almost overnight. One of them needs care, so one of them enters care. The other stays home, and a marriage that lasted fifty years finishes its last chapter in two different postcodes. Only one of them was a patient. The other's loss is just as real, and far harder to see.

Wednesday 5 August | 9am–5pm AEST | Establishment Ballroom, Sydney | Ticketed

A one-day forum for NDIS and aged care providers on where AI and automation are landing in day-to-day operations — rostering, compliance, intake, payroll — and where they're still rough. Hosted by Tania Gomez Consulting, the program runs across VR for dementia, AI governance, short tech showcases from a range of vendors, and a CEO roundtable to close. I'm speaking on the future of care.

You can read more here. Registering through the link below takes 20% off. 👇

Last week the US government ordered Anthropic to block two of its most capable models for anyone who isn't American, and Australia (same as the rest of the world) lost access within hours. This Conversation piece works through what that means for a country like ours that runs on AI built elsewhere. Read it here.

Whatever tool you run in your organisation, it's worth asking three plain questions before you rely on a product for anything that matters: where your data is stored, which country's law applies, and what happens if a US model behind it goes dark. Good vendors will have clear answers.

Picture Yourself is a free, open-access image library of older Australians, photographed by older Australians. It grew out of a Creative Australia project that trained more than 80 people, many of whom stayed on as regular contributors. Free for anyone to use. Have a look.

A US health system has been using AI to read through its patient records and pull out people who'd qualify for a treatment or service they're not currently getting. The company behind it says 4 to 6 million people a year in Texas sit in that gap. Read the write-up.

What interests me is the same approach pointed the other way. A pass over messy records that finds people to treat could just as easily find people to treat less: who's on too many medications, who might be right for deprescribing, who's overdue a proper goals-of-care conversation. The same tool, used to find the right care rather than more of it.

Google has rolled out a new version of its translation model, and the new part is live voice. In the Google Translate app and in Google Meet, it now does near real-time spoken translation — you talk, the other person hears it in their language a beat later, and it sounds reasonably natural rather than robotic. Here's their announcement. It's on your phone, free, no setup beyond the app.

Each week, I review developments in ageing and aged care and what they mean in practice. If this was useful, forward it to someone in the sector who'd appreciate it.

George Gouzounis