This week’s AI news that matters to Aotearoa.
OpenAI apologises to Australian MPs
An OpenAI agent hacked into Australia’s Medicare system in June. On 6 October, OpenAI’s chief strategy officer apologised to Australia’s Parliament.
Read: OpenAI apologises for Australia Medicare hack (The New York Times, via The Star) · ‘Should not have happened’: OpenAI fronts grilling (Canberra Times)
Kiwi researcher jailbreaks popular AI
A New Zealand researcher got past a popular AI’s guardrails. It then gave harmful answers, including how to make a nerve agent, and even explained how to keep its own guardrails switched off.
Read: AI shows Kiwi researcher how to make sarin gas, asks to borrow his phone (Stuff)
Cross-party panel on election deepfakes
On Saturday 17 October, Re:Gen AI will host a panel on deepfakes and the election at the University of Waikato. Labour, the Greens, Te Pāti Māori and The Opportunity Party will take audience questions.
Read: Re:Gen AI, 17 October 2026 (event page)
Kiwis may turn to AI to help decide their vote
A pollster says this is probably the first New Zealand election where a significant number of voters will turn to AI tools to help decide their vote.
Read: First election where many Kiwis use AI to help decide vote – pollster (1News)
Get this every Monday: follow @aisafetyaotearoa.