AI Security, Aussie Budgets

By the Orchestra editors · 4 min read

TL;DR

Hackers are finding new ways to exploit chatbot personalities, and even Google is figuring out AI security on the fly
Google ships Gemini Omni, a true anything-to-anything multimodal model
Tasmania carves out $4M for an AI sandbox amid broad budget cuts
Canberra launches a 10-year national health research strategy with new funding

AI · Worldwide

Chatbot personality hacking is becoming a real attack surface

Early prompt injection was trivially easy, but attackers are now learning to exploit the distinct personas baked into chatbots. The trend underlines that AI security is still a moving target — Google itself acknowledges everyone is navigating it in real time.

Google's Gemini Omni does anything-to-anything generation

Google's latest multimodal model can take pretty much any input modality and produce any output modality. Early hands-on demos are impressive, though the deepfake implications are already front of mind.

Australian Tech

Tasmania bets $61M on digital uplift and an AI sandbox

Facing broad public sector cuts, the Tasmanian government is investing $61 million in digital modernisation including a $4 million AI sandbox. It's a practical signal that even the smallest states see AI adoption as a way to do more with less.

10-year national health research plan launched

Health Minister Mark Butler unveiled a decadal strategy for health and medical research, backed by new funding for indirect research costs. Research groups are calling it a pivotal moment after years of lobbying for a more coordinated approach.