Statement of AI Risk
Question
The National AI Plan commits the Government to ''taking proactive steps to prepare for any potential AI-related incident'', with the AI Safety Institute explicitly identified in the Plan's ''Keep Australians Safe'' section as part of how this commitment will be delivered. Does ''any potential AI-related incident'' include loss of control of advanced AI systems, AI-enabled CBRN catastrophes, and other catastrophic and existential-level risks? What proportion of AISI's 2026-27 work programme, by FTE or budget, is allocated to evaluating frontier models for these risks? The National AI Plan references the Bletchley Declaration, which states there is ''potential for serious, even catastrophic, harm'' from frontier AI, and identifies ''unintended issues of control relating to alignment with human intent'' as a substantial risk. The National AI Plan itself does not use the words ''catastrophic'' nor ''existential'', and only uses ''alignment'' in the non-technical sense. What advice did the Department provide on whether the National AI Plan should adopt the Bletchley framing? Was the omission deliberate? Who made the decision to omit 'catastrophic' and 'existential' from the Plan? The 2023 Statement on AI Risk, signed by leaders of OpenAI, Anthropic, Google DeepMind, and hundreds of senior AI researchers, stated that ''mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks''. Has the department or any other Commonwealth agency formed a view on the Statement on AI Risk? What proportion of the AISI's mandate addresses extinction-level risk specifically? Is AISI a member of the International Network of AI Safety Institutes convened at the Seoul AI Safety Summit in May 2024? If yes, when did it join and what working groups does it participate in? If not, why not, and what is the timeline for joining?
Answer
See attached response.
Comments · 0
Please log in to post a comment.
Loading comments...