Extract from ABC News
Kate Conroy has been described as a "global expert in responsible AI". (Screenshot: Royal Australian Air Force)
In short:
Kate Conroy has been chosen as the inaugural general manager of Australia's AI Safety Institute.
The institute will sit inside the Department of Industry, rather than operate as a standalone independent regulator.
What's next?
The institute's early work will test how far a small new body can shape Australia's response to AI risks.
A philosopher and Royal Australian Air Force reservist has been chosen to lead the federal government's efforts to keep Australians safe from the harms of artificial intelligence.
Two sources with knowledge of the appointment but not authorised to speak publicly told the ABC Kate Conroy started as the inaugural general manager of the Australian AI Safety Institute earlier this month.
The institute is a new government body that sits within the Department of Industry, Science and Resources. It is not a standalone, independent regulator.
Its job is to monitor, test and share information on emerging AI capabilities, risks and harms.
A spokesperson for the Department of Industry, Science and Resources confirmed the appointment.
"Dr Conroy has been appointed as general manager, AI governance. This team is working to establish the Australian AI Safety Institute," they said.
Dr Conroy takes up the role as Australia works through how to regulate AI, with the technology already integrated into workplaces, national security, education and public services.
From philosophy to inside a Poseidon P-8A
The department spokesperson described Dr Conroy as a "global expert in responsible AI".
Dr Conroy comes to the role with a background of working with governments on AI ethics and risks, particularly around autonomous systems and robotics.
She remains the lead of responsible AI at the Royal Australian Air Force, where she has been involved in the introduction of its first AI system to run live inside a plane's systems, the Poseidon P-8A, as part of AUKUS Pillar II.
Kate Conroy has previously worked with governments on AI ethics and risks. (LinkedIn: Kate Conroy)
In a Department of Defence media release, Dr Conroy described the plane's AI system working as "a helper, not a decider".
She said it would present incoming data and give the crew feedback on the plane's operation but left the decision-making to the humans.
In 2022, Dr Conroy authored an academic paper warning militaries could "cook" personnel into "bad apples" if they delegated more parts of the decision-making process in war situations to AI.
Dr Conroy has also worked as the chief scientist for the Trusted Autonomous Systems Defence Cooperative Research Centre, where she co-authored technical reports and created toolkits for the ethical use of AI in armed forces.
Outside of the military, Dr Conroy had primarily worked with the government and as an academic.
Immediately before her role in the air force, Dr Conroy worked for the Queensland government, introducing rules to evaluate and manage the risks of AI in government.
She co-authored a framework for governments using AI that was adopted as the national standard in 2024.
She obtained a PhD in philosophy in 2013 and has held roles at the University of Queensland and Queensland University of Technology.
Australia's approach to AI safety
Dr Conroy's appointment puts an AI ethics specialist, rather than a conventional technical specialist or former regulator, at the head of Australia's first AI safety body.
She is the first confirmed member of the AI Safety Institute, which was announced in November 2025 as part of the rollout of the national AI plan.
Minister for Industry, Innovation and Science Tim Ayres said at the time that Australia would only be able to seize the opportunities of AI if it was also able to manage its risks.
The AI Safety Institute would be used to assess these risks, Mr Ayres said.
"The institute will be the government's hub of AI safety expertise. It will operate with transparency, responsiveness and technical rigour to make sure Australians are confident to use this game-changing technology safely," he said.
The institute has been given funding of $29.9 million over four years and was set to be established in "early 2026".
Its creation followed the government's decision to move away from a proposed single AI Act, a central AI regulator and advisory body of experts, and instead rely on existing regulators and agencies to manage AI harms across their own areas.
Under the national AI plan, the institute will have to work on both "upstream" risks and "downstream" harms.
The plan defines upstream risks as the capabilities of AI models and the way they are built and trained.
Downstream harms are the real-world effects people experience when AI systems are used.
That gives the institute a wide remit: from advanced model testing and international AI safety research to the everyday consequences of AI systems being used in workplaces, government services, finance and consumer products.
Australians using AI but do not trust it
The institute's imminent opening comes at an important time for AI in Australia.
Since the generative AI boom was sparked with the release of ChatGPT at the end of 2022, there has been a rapid growth in use of products and tools powered by the new technology.
At the same time, public polling consistently shows there is a broad distrust of the technology among Australians.
A recent global survey of attitudes towards AI in 23 countries found Australians were already using AI widely but had the equal-worst sentiment towards the technology.
The study found 68 per cent of Australians were worried about losing control over decisions AI made on their behalf, while 81 per cent supported stronger rules for how organisations used AI.
There has also been public discussion over which safety risks should be the focus of the safety institute's modest resources and capacity.
Toby Walsh says the institute will need to carefully decide what to focus on, given its modest resources. (ABC News: John Gunn)
The University of New South Wales AI Institute's chief scientist and leading AI researcher, Toby Walsh, said the new body would need to carefully decide what to focus on.
"The big challenge for the AI Safety Institute is being able to provide meaningful support and advice given the level of funding it's got," he said.
Some experts want Australia to focus heavily on frontier models and their potential future risks: the most advanced systems developed by companies such as OpenAI, Anthropic and Google DeepMind and what they might be capable of in the years to come.
A significant amount of work by the long-running international AI safety community has been focused on so-called "existential risks" to humanity, including rogue AIs, bioweapons and other doomsday scenarios.
Others argue the institute should focus on the current harms, including discrimination, privacy breaches, scams, unsafe products and opaque decision-making in government and business.
Its sister institute, the UK's AI Security Institute, has emerged as one of the world's leading safety organisations.
It was one of the first organisations, and the first non-US-based one, to publish a public evaluation of Anthropic's Claude Mythos, a new AI model which was held back from public release by its maker for being too "dangerous".
Australia has already moved to connect its institute to the UK's. Earlier this week, the governments signed a memorandum of understanding on safe and secure AI.
The agreement covers sharing information about emerging AI capabilities and risks, joint research on measuring and testing AI systems, and support for international AI measurement and evaluation work.
The Department of Industry spokesperson did not say when the Australian AI Safety Institute would formally open, only that it would be "operational soon".
No comments:
Post a Comment