An independent Australian convening
Deciding under uncertainty, before the noise decides for us: Australia's decisions on frontier AI
The public conversation about advanced AI has become loud, polarised and hard to act on. That is exactly why this convening exists: to get people who disagree into the same room, working from the same facts, and put something on paper our decision makers can actually use.
Early stage · participants and venue being confirmedWhy now
Global existential risk from advanced AI has become one of the loudest topics in public life, and one of the hardest to discuss well. The volume itself has made careful, cross-partisan conversation harder to have, not easier, at precisely the moment Australia is building the institutions that will govern this.
That is the problem this convening is built to solve, not a reason to wait for a quieter moment. Under conditions of real uncertainty about how the technology and the politics will move, the answer is not silence. It is getting people who see this differently, across the political spectrum, into one room and asking them to write down what we actually know, don't know, and need to decide.
Australia is building its AI institutions in real time: a national AI Safety Institute, a new Office of AI, a parliamentary inquiry and a state royal commission. Each is working on part of this picture, but none of them is set up to hold this exact conversation. This is a working session, not a talkfest: its purpose is to gather the specific information Australia's decision makers need, and hand it to them while there is still time to use it.
The workshop
Target number of experts, spanning both sides of the political spectrum, invited close to the meeting date so the room fits the questions.
A structured, expert-led discussion with decision makers in the room, not only receiving the output.
Months ahead: the decisions Australia must make, and by when.
Three areas of focus
Deliberately narrow, and deliberately not aligned to any one camp in the public debate. The workshop is limited to the long-range and catastrophic questions Australia will need answers to, currently absent from Australian processes, examined by people who don't agree with each other going in.
Loss of control and misalignment
What should Australia require, test or prepare for as AI systems become more capable and more autonomous?
Catastrophic malign use
How should Australia think about AI-enabled biological, chemical and cyber threats, and its place in international controls?
Strategic and military stability
What independent voice can Australia have on AI and strategic stability among allies and in the region?
What comes out of it
The outputs are a one-page map of information our decision makers need over the next 12–24 months to answer upcoming questions on frontier AI governance, and a discussion report by the end of January 2027.
A one-page information map
What decision makers need to know over the next 12–24 months to answer upcoming questions on frontier AI governance.
A discussion report
The full record of the convening, published by the end of January 2027.
How it works
Independent. Run separately from any government agency, company or advocacy group, so the output can be trusted by all of them.
Cross-spectrum by design. Participants are chosen so the room includes real disagreement, not a single point of view restated twelve times.
Built for uncertainty. We don't wait for consensus on the future to act. The goal is a shared, honest account of what is known, contested and unknown, useful now.
Decision-focused. Every session is tied to a real decision with a real owner and a real deadline.
Convenor
I'm not a career policy official, and I don't pretend to be. My background is strategy and delivering complicated multi-party work: strategy at CSIRO, four years at the Australian Embassy in Berlin working with state governments and briefing ministers, then delivering the German Red Cross Anticipation Hub's global dialogue platform, which brought together 300+ participants. I've followed AI safety since 2019. I started this because many of the Australians who know most about frontier AI risk work overseas, and the people making Australia's decisions rarely hear from them in one room. I'm not the expert on the answers. My job is to get the right people in the room and make sure what comes out is usable.
Jessie Kelly · Berlin · LinkedIn
Status and contact
Planning stage. Host, venue, dates and participants are still being confirmed.
No organisation is named as a host, funder or endorser on this page until that has been agreed in writing.