AI news
What's happening in AI
The AI news worth knowing, in plain English. New models, big moves, safety, and what it all means for business.
Summarised by AI, read by Andrew. Each night AI picks the most interesting AI stories and summarises them from the original reporting. I read every edition and correct anything that's wrong, and every story links to its source. How I use AI
Thursday 15 October 2026
7 stories
New models
Google has released Gemini 4 Argon, its new top model, in the Gemini app and in AI Mode in Search. For now it's only for paying AI Pro and Ultra subscribers; in the US those plans cost US$20 and about US$100 a month. Developers get introductory pricing of US$2 per million input tokens (the chunks of text AI firms bill by) until 30 November, when it doubles. Benchmarking firm Artificial Analysis says Argon matches OpenAI's GPT-6 Astra on its overall index at about 60% of the cost per task.
Why it matters: If you already pay for Google's AI plan, you can try it now on proposals, spreadsheets and documents, with nothing new to sign up for.
Source: Ars Technica · Also reported by Google, Engadget
Safety & policy
Eight US states, including Iowa, Kansas and Arizona, each sued OpenAI in their own courts on 14 October (US time). They allege OpenAI broke consumer-protection laws by marketing ChatGPT as safe for young people while it harmed vulnerable users, and that it collected data from minors. The states want penalties and changes to how ChatGPT works for young users. OpenAI says it has built strong age-specific protections, including parental controls, and wants rules that cover the whole industry. If this raises concerns for you, call Lifeline on 13 11 14.
Why it matters: If you put a chatbot in front of customers, consumer law covers what you claim about it. Don't over-promise on safety or accuracy, and keep a person in the loop.
Source: The Verge · Also reported by The Guardian, CNBC
Business
Online retailer Kogan now has an AI agent handling customer service, built on Salesforce's Agentforce. Speaking at Salesforce's Dreamforce conference, CTO Goran Stefkovski said Kogan's old rules-based bot resolved only about 6% of enquiries. The rebuilt agent went live in about six weeks and now covers 10 jobs, from order tracking (around 60% of inbound requests) to product recommendations. It can even spot the same damaged-product photo being reused across claims.
Why it matters: If one question dominates your inbox, that's your first AI project. Connect it to your order system, start before it's perfect, then expand.
Source: iTnews
Industry
Character.AI, the AI companion chatbot app, has filed for Chapter 11 bankruptcy in Delaware, blaming high development costs plus legal and regulatory pressure. Disney has made an opening bid of US$20 million for a non-exclusive, perpetual licence to some of its technology. The app stays online during the sale, which should wrap up by late November or early December, and it could be wound down if no buyer emerges. Its privacy policy allows user information, including chat logs, to be transferred in an asset sale.
Why it matters: AI start-ups can fold fast. Before you rely on a small AI vendor, check how you'd export your data, and what happens to it if they're sold.
Source: Ars Technica · Also reported by CNBC, Engadget
Applications
Google Messages is getting an opt-in "enhanced protection" setting that uses Gemini to judge whether a chat from an unknown sender looks like a scam. Google says messages stay encrypted and are analysed in a secure space it can't see, and an on-device model checks unfamiliar links before you tap. It rolls out over the coming weeks, starting with Pixel 10 and newer, Samsung's Galaxy S26 series and the latest Galaxy Z Flip and Fold. The reports don't say when it reaches Australia.
Why it matters: Scam messages to staff phones are a real business risk. Check which of your team's phones will get the new protection, and switch it on when it arrives.
Source: Engadget · Also reported by The Verge, Google DeepMind
Safety & policy
An ABC investigation found AI-generated and AI-altered images in property listings around Australia, from virtually staged rooms to invented families. The ACCC says AI images mustn't leave an overall false impression, because misleading conduct is banned under Australian Consumer Law. Queensland's Office of Fair Trading published a best-practice guide in June urging agents to disclose AI use, and has had six complaints about digitally altered listing images since January 2025. The Real Estate Institute of Australia backs AI for showing a property's potential, as long as it's disclosed.
Why it matters: The same rule applies to any business using AI images in marketing, from products to venues to food. Disclose the edits and don't create a false impression.
Source: ABC News
Gaming & entertainment
YouTube Music has started labelling tracks where AI played a significant role, and tapping the label explains how it was used. Music partners declare whether a track is fully AI-generated, partly AI-generated or not AI, and YouTube runs its own detection too. AI involvement alone won't stop a track earning money. Separately, Deezer says a switch to hide AI-generated tracks, albums and artists arrives before the end of the year. In July it said about 90,000 tracks a day, more than half its uploads, were AI-made.
Why it matters: If you use AI-generated music in ads, videos or in-store playlists, expect labelling and disclosure to become normal, and check the licence terms.
Source: TechCrunch · Also reported by The Verge