If you’ve used a chatbot for anything more personal than a recipe search, you’ve probably hit the same wall: you explain something once, close the tab, and a few days later you’re explaining it all over again. Worse, most of us have no real idea where that conversation lives once we’ve had it – which server, which country, under which laws.
That question has been getting louder lately. As AI tools quietly become part of how women plan their weeks, manage their health, run their small businesses and even process hard days, the conversation has shifted from “what can AI do?” to “who’s actually looking after my data, and can I trust the people building this?” It’s a fair question. Most of the AI we use every day – the chat assistants, the recommendation engines, the tools baked into our phones – is designed, hosted and governed somewhere else entirely, under someone else’s rules.
That’s the gap a new Adelaide-based company called Metacognition wants to close.
Who’s Behind It
Metacognition was founded by three researchers who’ve spent years at the pointy end of Australian AI: Professor Anton van den Hengel, Professor Stephen Gould and Dr Paul Dalby. Between them, they’ve built research institutes, trained the next generation of AI talent, and worked on projects that have generated real, measurable industry value – not just theory.
They’ve just come out of stealth at Deep Tech Week 2026 in Sydney, backed by Main Sequence Ventures (a serious name in Australian venture capital) and the South Australian Government, which has also been investing in the state’s broader AI research capability. It’s early days – the company hasn’t launched its first product yet – but the backing and pedigree behind it are already turning heads.
The Problem They’re Trying to Solve
Here’s the everyday frustration Metacognition is chasing: today’s AI tools essentially forget you. Every conversation starts from close to zero. You can tell a chatbot exactly how you like things done, but within days it forgets you ever said it. There’s no real thread of understanding building over time – no sense that the tool is getting to know you, your context, your life.
Metacognition’s founders describe the alternative as something closer to working with a trusted colleague – one who remembers the shorthand, understands the nuance, and gets more useful the longer you work together, rather than resetting every session. It’s a genuinely relatable frustration, and one plenty of us have felt without necessarily having the words for it.
Why the “Australian” Part Actually Matters
This is where the story gets interesting beyond the tech specs. Right now, overseas companies – mostly in the US – build, host and legally govern the vast majority of AI most Australians rely on: the big chat assistants, the tools embedded in our apps. That means the data, the design choices, and the safety standards applied to those systems all sit under someone else’s jurisdiction.
Metacognition is pitching itself as a local alternative: built in Adelaide, employing Australian talent (and drawing global talent here too), and – notably – built under Australian rules. The founders have said they’re adopting Australia’s voluntary AI Safety Standard as their starting point, not their ceiling, with internal benchmarks designed to go further. They’ve also spoken specifically about keeping humans in control at key decision points, rather than designing systems that quietly make consequential calls on their own.
In an environment where “AI safety” can feel like a vague marketing line, that’s a meaningfully specific commitment: a local company, under local law, aiming to exceed a local safety benchmark rather than simply meeting the minimum.
Why This Should Matter to You, Not Just the Tech Crowd
It’s easy for AI sovereignty conversations to stay stuck in policy-speak – data centres, jurisdictions, VC funding rounds. But strip that back and the real question is a personal one: when you type something into an AI tool, who is that tool actually accountable to, and does it treat you like a whole person or a string of disconnected prompts?
For women in particular, who are often navigating AI tools around genuinely sensitive territory – health tracking, financial planning, career moves, family logistics- the question of who’s behind the system, and under what rules it operates, isn’t a niche technical detail. It’s the difference between a tool that quietly works for you and one you have to manage carefully, prompt by prompt, hoping it doesn’t lose the thread – or your trust.
Metacognition’s first product is still in development, so we don’t yet know exactly how (or when) it’ll show up in the tools we use day to day. But the intent behind it – AI that remembers responsibly, reasons more like a thoughtful collaborator, and answers to Australian safety standards rather than someone else’s – is a promising signal in a space that badly needs more of them.
We’ll be watching this one closely, and we’ll bring you the details the moment there’s more to share.
