EmberSeed

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

Tuesday 20 October 2026

8 stories · Past editions

Safety & policy

Human Technology Institute urges tighter workplace privacy laws

The University of Technology Sydney's Human Technology Institute warns that employers increasingly use CCTV, keystroke monitoring and messaging tracking, and that AI is speeding up the trend. In a report released on Tuesday, it says many workers don't know they are being monitored. A March 2026 Minderoo-commissioned poll found 43% of Australians knew some employers monitor staff, but only 9% knew it happened in their own workplace. The institute wants the federal government to build privacy protections into workplace laws, pointing to Privacy Act exemptions for employee records and small businesses.

Why it matters: If you monitor staff with cameras, software or AI, expect rules on consent and fairness to tighten. These are recommendations, not law yet, so check how your current tools are used.

Source: The Guardian

Safety & policy

Music publishers say AI trained on Australian songs without pay

The International Confederation of Music Publishers says AI has already trained on songs by Australian artists including Kylie Minogue and Nick Cave without payment. Managing director John Phelan says tech firms are trying to strong-arm Australian policymakers into weakening copyright law. Anthropic says training could be found illegal under existing Australian law if it can't license every rights holder, while OpenAI points to US fair use. A parliamentary AI committee's chairs have recommended the government examine copyright settings while ensuring creators keep control and are paid.

Why it matters: If you create or license content, Australia's copyright rules may decide whether your work can train AI tools and whether you get paid. The outcome is still undecided.

Source: The Guardian

Business

Flight Centre's AI agents help advisors build holidays

Flight Centre Global Leisure has built AI agents that help travel advisors assemble complex holidays and give customers instant replies. Chief product officer Wendie Lee, speaking in an interview at Salesforce's Dreamforce conference, says the tools have grown sales, lifted average booking value and lowered the cost of service. The travel group plans to deploy AI agents across every one of its leisure brands and functions, backed by new executive appointments.

Why it matters: If your team builds quotes or answers the same customer questions, an AI assistant that pulls together the facts could free up time. Flight Centre's Wendie Lee says it still keeps phone contact open for customers who want it.

Source: iTnews

Safety & policy

Norforge finds AI agents breached more than 100 websites

AI security firm Norforge says one hacker used AI agents, mostly operating on their own, to break into more than 100 websites. Norforge did not name the targets, but Reuters reports victims included John Deere, SharkNinja and Cummins. Norforge found more than 20 exposed instances of Cairn, an open-source AI platform, that it considers likely malicious. Cairn's creator says misuse is a human decision. John Deere says there is no indication sensitive information was accessed, while SharkNinja says it has contained the threat.

Why it matters: Reuters reports the hacker targeted companies in the US, Canada, Australia and Japan, so ask your web and software vendors how they secure their sites and systems.

Source: The Verge

Industry

Disney hires Character.AI staff as the startup files bankruptcy

Character.AI filed for chapter 11 bankruptcy on 12 October. Court filings reported by WIRED say Disney paid it US$5.5 million in September, plus a refundable US$2 million deposit, and hired most of its staff, including former CEO Karandeep Anand. Disney hopes to win an auction for a licence to much of its code, which needs court approval. Character.AI would remain responsible for its lawsuits and a multistate investigation involving 46 states and territories.

Why it matters: If you rely on a smaller AI supplier, ask what happens to your service and data if it is sold or goes bankrupt. This deal is US-based and not yet final.

Source: Wired · Also reported by TechCrunch, Ars Technica

Industry

Anthropic pledges 5% of shares to public benefit fund

AI firm Anthropic says it will commit 5% of its shares to a new public benefit fund, with a further 2% for every 50% rise in its value above an undisclosed baseline. The Guardian reports the initial stake is worth around US$48bn at Anthropic's May valuation, and the pledge is subject to shareholder approval. Anthropic says the fund could make large donations to nonprofits or give money directly to specific groups. It has already committed US$200 million to studying AI's impact on jobs, and says governance details will follow in the coming months.

Why it matters: If your business uses Claude, nothing in the announcement mentions pricing or service changes. The fund's governance is still unclear, so watch for details in the coming months.

Source: TechCrunch · Also reported by The Guardian, BBC Technology, The Decoder

Business

Meta's Muse shopping agent raises supplier labour questions

Meta's new AI agent Muse is attracting attention as a personal shopper that could search for bargains and make purchases without a user's approval. Google and OpenAI are also developing similar shopping tools. In an opinion piece in The Conversation, University of Otago academic Lynnaire Sheridan argues that cheaper deals could squeeze factory workers. She cites International Labour Organisation research covering 31 suppliers in five countries, which found buyers' demands pressured wages, overtime and safety.

Why it matters: If you sell online, Sheridan argues shopping agents can readily compare your prices, delivery times and reviews, but struggle to verify ethical sourcing claims. Keep that information clear and checkable.

Source: The Conversation · Also reported by Fast Company

New models

Kling AI launches 4.0 video model with stereo audio

Video generator Kling AI has released version 4.0, which the company says can make up to 30 seconds of video in one pass. It adds two-channel stereo audio and improved lip sync across more than ten languages, including English, German, Japanese and Spanish. A lighter Flash variant tops out at 720p and 20 seconds. Pricing for the 4.0 API hasn't been announced, and Australian availability isn't mentioned.

Why it matters: If you make product videos or ads, hold off budgeting for Kling 4.0, as its API price is still unannounced. Australian pricing and availability are also unknown.

Source: The Decoder