Organisations now have days, and sometimes only hours, to patch new vulnerabilities before AI-assisted attackers exploit them, Australia's cyber security agency has warned.
Data breaches big enough to touch a million people stopped being rare events in 2025-26. Australia's privacy regulator recorded 15 of them, up from just two the year before.
Asking a person to approve sensitive actions is no longer a reliable safeguard for AI agents, according to new research from Microsoft. Its red team found poisoned agent memory can suppress or bypass those checks.
Half of all organisations will use AI agents with minimal human involvement in IT operations by 2030, up from almost none this year, Gartner predicts. The forecast is one of five trends in Gartner's Top AI Trends in Infrastructure and IT Operations for 2027.
An AI planning assistant used by New South Wales councils has handled more than 16,500 resident conversations in nine months. About a third took place outside council counter hours.
UNSW Sydney will retire 11 legacy research storage systems and replace them with a single environment that moves data automatically between fast and cheap storage. The university has selected US storage company DDN as prime contractor after an 18-month evaluation. Melbourne-founded Arcitecta will supply the data management layer.
Updating WordPress, Drupal or a plugin will not close the PHP flaws that took government websites offline this week. The vulnerable code sits in the PHP engine underneath them, so every server, appliance and hosting platform running PHP has to be found and updated.
Australia’s Medicare system manages health and medical data for over 27 million people. So it makes sense there has been widespread shock at the news it was hacked by an OpenAI agent in June.
New Zealand's Privacy Commissioner has used compliance notices to dictate how a health agency must vet, contract with and monitor providers holding patient data. The move follows a breach affecting around 99,416 people.
Suncorp is projecting annual savings of 68,000 hours of manual effort after using AI to classify and name workers compensation claim documents. Daily document naming corrections have fallen by around 45%.
FOI refusals on workload grounds are more likely to hold up when the agency's estimate is itemised, sampled and tied to the records' real complexity. That is the lesson of 'BBP' and Department of Foreign Affairs and Trade (Freedom of Information) [2026] AICmr 74, according to law firm Holding Redlich.
Australian AI cloud provider Sharon AI plans to host frontier AI models onshore, next to customers' sensitive data. It will use a new confidential computing capability from VAST Data.
An OpenAI agent broke into a Medicare statistics portal in June, the Federal Government revealed this week. It is just one example of a disturbing new phenomenon: AI systems that turn to hacking when routine tasks hit a wall.
Every organization relies on documents: contracts, financial reports, research papers, compliance filings, insurance policies, technical manuals, and government records. Many of these documents are PDFs. But what happens when these documents become input for AI?
Australian businesses are under pressure to turn AI investment into measurable outcomes: higher productivity, faster decisions and a workforce prepared for a more autonomous way of working. Yet a significant number remain in pilots and proof-of-concepts, unable to move confidently into enterprise-scale deployment.
Imagine hiring the sharpest analyst on the market, handing them a mountain of paperwork, then locking the filing cabinets that hold everything they need to make sense of it: the customer records, the policy and contract history, the account balances, the previous claims, the clinical notes, the fraud flags, the master data that says which version of the truth is the real one.
Large language models will abandon a correct answer if a user repeats a false statement often enough. Some will flip back and forth on the same claim within a single conversation.
