AI moves every week. The Radar picks the handful of changes that matter to Australian businesses and professionals, explains them in plain English, and gives each one a verdict, so you know whether to act, try it, watch or ignore it.
Reviewed 29 September 2026 · next review by 6 October 2026 · every item links to its source.
From 10 December 2026, organisations covered by the Australian Privacy Principles must explain in their privacy policy any substantially automated decisions that significantly affect people.
What it means for you: List where AI makes or shapes decisions about customers or staff (credit, pricing, hiring, eligibility) and update your privacy policy before the deadline.
Anthropic released Claude Sonnet 5.5. It runs 30%+ faster and costs up to 30% less per task than Sonnet 5, and nearly matches Opus 5.5 on knowledge-work benchmarks.
What it means for you: For most everyday work (drafting, analysis, customer replies, agents) this is now the price-to-quality sweet spot. If you pay for a top-tier model for routine tasks, test this one.
A Senate inquiry asked the OpenAI and Anthropic CEOs to appear, after reports that an OpenAI agent accessed data on the government’s Medicare portal. OpenAI said it found no evidence that patient records were accessed. Both companies declined the 1 October hearing, citing short notice.
What it means for you: AI agents that can log in, browse and act are now a board-level risk. Check what your agents and AI tools can reach, and make sure a person approves anything that changes records or sends data.
Trackers counted more than 20 new AI models from over a dozen providers in September alone.
What it means for you: Don’t switch tools every time a model launches. Keep a small test set of your real tasks and re-check your choice each quarter, or when prices move by 30% or more.
GPT-6 Sol costs US$2 / US$10 per million input/output tokens and GPT-6 Luna costs US$0.10 / US$0.50: about half the price of the GPT-5.6 models they replace. OpenAI says these are permanent prices.
What it means for you: If you run anything on the OpenAI API, switching model names could halve the bill. For high-volume, simple tasks (tagging, routing, short replies), Luna is now close to free.
Anthropic released Claude Opus 5.5 at US$4 / US$20 per million tokens, 20% below Opus 5. Anthropic says it matches Claude Fable 5.1 on most work and costs about 40% less to run than Opus 5 on typical workloads.
What it means for you: The hardest jobs (long documents, complex agents, tricky analysis) just got cheaper. Keep it for the steps where cheaper models fail your tests.
Qwen3.8 (including a 27B model under the Apache 2.0 licence) and DeepSeek V4 (MIT licence) are the current open-weight families. September leaderboards ranked GLM-5.3 and Kimi K3 highest among open-weight models.
What it means for you: Private AI that runs on your own hardware, with no per-token bill, is now realistic for focused jobs like triage, extraction and Q&A over your own documents. Check each model’s licence first.
Google released Gemini 3.8 Flash, with big gains in coding and agentic tasks over 3.7 Flash, at an introductory price that runs until 31 December.
What it means for you: A strong, cheap option for high-volume work, especially if you already use Google Workspace. Lock in a test before the introductory price ends.
The Government announced Australian Standards for AI and set up an Office of AI in the Department of the Prime Minister and Cabinet (operating from 15 July). Legislation is expected in early 2027.
What it means for you: Nothing is mandatory yet, but the direction is clear. Organisations with an AI register, named owners and human oversight will have the least rework when the laws land.
Guess our verdict on each headline before you read it. How good is your AI hype detector?
Standard API prices in US dollars per million tokens (roughly 750,000 words). Chat subscriptions are priced per seat instead. Prices change often, so check the provider before you commit.
| Model | Tier | Input | Output | Best for |
|---|---|---|---|---|
| Claude Opus 5.5Anthropic | Frontier | US$4.00 | US$20.00 | Hardest reasoning, long documents, complex agents |
| Claude Sonnet 5.5Anthropic | Everyday | US$2.00 | US$10.00 | Everyday drafting, analysis and agents |
| GPT-6 SolOpenAI | Everyday | US$2.00 | US$10.00 | Everyday drafting, analysis and agents |
| Claude Haiku 4.5Anthropic | Budget | US$1.00 | US$5.00 | Fast, simple, high-volume tasks |
| Gemini 3.8 FlashGoogle | Budget | US$0.75 | US$3.75 | High-volume work, coding, Google WorkspaceIntroductory price until 31 Dec 2026 |
| GPT-6 LunaOpenAI | Budget | US$0.10 | US$0.50 | Tagging, routing, short replies at huge volume |
| Qwen3.8 27BAlibaba (open-weight) | Open | Free to run yourselfYou pay for hardware | Private AI on your own hardwareApache 2.0 licence | |
| DeepSeek V4 FlashDeepSeek (open-weight) | Open | Free to run yourselfYou pay for hardware | Private AI on a small GPU serverMIT licence | |
Pick a task and a monthly volume to compare monthly API costs across models. These are estimates: real token counts vary with your prompts and documents.
Updated with every news run (Tuesday, Thursday and Saturday).
One email: what changed, what it means, what to do. Free, and you can unsubscribe any time.
Not sure which tools fit you? Find your AI stack in 60 seconds →