Algorithmikē Psychē
If Q3 2025 was about people quietly admitting how much they didn't know yet, Q4 was the quarter that turned into action, and into numbers. Australia stood up a new institute to actually test these systems. A survey put a real figure on how many people are already using AI this way. And a UN-linked summit put mental health access on the agenda for the parts of the world that have the least of it.
One new Australian institution, one number worth sitting with, and one global gathering that widened the conversation past the countries that usually dominate it.
A New Institute, With a Number Attached
The Australian Government announced it would set up an Australian AI Safety Institute, backed by $29.9 million in funding, to test frontier AI models and act as a central point where government, regulators and the public can get a straight answer about what these systems can and can't be trusted to do.
It's worth being clear about what this is and isn't. It's not a new regulator with new powers — existing bodies like APRA, ASIC, ACCC and the OAIC stay in charge of enforcement. What Australia has built instead is a technical hub: a place whose whole job is to actually understand the technology well enough to tell everyone else the truth about it. Given how much of this year's global conversation has been about institutions admitting they're behind, having somewhere dedicated to catching up is not a small thing.
Read the announcement →Ayres, T. (2025, November 25). Establishment of Australian AI Safety Institute [Media release]. Australian Government Department of Industry, Science and Resources.
One in Four Americans Turns to AI for Mental Health Information
A nationally representative survey of more than 5,500 US adults found that roughly 24 percent had used an AI tool to explore a mental health or emotional concern — rising to nearly a third among people specifically seeking help for something they were struggling with. That's not a niche behaviour anymore. That's a meaningful share of a country quietly turning to AI before, after, or instead of seeing a doctor.
What the survey doesn't answer is the part that matters most: what guidance are these tools actually giving, and to whom? There's no standard yet for what a "good" AI response to someone in distress looks like, and no way to know how often people are getting one. The number itself is the finding. What it means is still being worked out.
Read the findings →West Health & Gallup. (2025). Millions of Americans now consult AI before, after, and sometimes instead of seeing a doctor.
Mental Health Gets a Seat at the Global AI Table
At the AI for Good Global Summit in Geneva — the UN's own venue for this conversation — mental health had its own dedicated session for what may be the first time at this scale. The focus wasn't the usual worry about chatbots going wrong. It was access: how AI might actually reach people in under-resourced regions who have never had a mental-health service anywhere near them.
It's easy to only hear about AI and mental health through stories about harm, and there's good reason those stories get told. But this session was a reminder that the same technology is also being discussed, seriously, as a way to reach people who currently have nothing. Both things are true at once, and this quarter's summit is where that tension was named out loud, on a global stage.
Read about the summit →ITU AI for Good. (2025, December 7). AI and the future of inclusive mental health care. AI for Good Global Summit.
From an institute built to catch up, to a number that made a quiet trend impossible to ignore, to a summit that widened who gets to be part of this conversation — Q4 2025 was the quarter the scaffolding started going up in more places than one. What it holds will be next year's story.
© 2025 The Algorithmikē Psychē — Q4 2025