New rules for AI-assisted hiring decisions, a landmark copyright payout, and the people building AI are now asking governments to help slow it down if needed.
G'day, today's stories span the legal, regulatory, and security side of AI, useful context whether you're an employer, a business owner worried about copyright, or just trying to keep tabs on where this is all heading.
In today’s briefing:
🇦🇺 Australia AI Watch: new AI decision-transparency rules for employers, plus a "Digital Duty of Care" on the way
🌐 1,100+ staff at OpenAI, Anthropic, Google, and Meta ask governments to help pace AI development
⚖️ Anthropic settles for $1.5 billion in the biggest AI copyright case yet, and faces backlash over how it sourced the books
🕵️ An AI security researcher found critical flaws in Bing before any human did
🛠️ Tool Spotlight: Perplexity for quick, sourced answers instead of a wall of search results
💡 Today's prompting tip: getting a quick copyright-risk check on your AI-assisted content
AUSTRALIA AI WATCH
🧾 New Transparency Rules for AI-Assisted Decisions Affecting Australians Take Effect from December
The Daily Rundown: From 10 December 2026, automated decision-making transparency requirements under the Privacy Act come into force, meaning businesses will need to be upfront when AI plays a role in decisions about customers or staff. Legal commentators expect this to matter most for employers using AI in recruitment, monitoring, rostering, or performance assessment, since those are the areas the government has flagged for closer attention, though the broader consumer side (things like automated pricing or eligibility decisions) is covered too.
The requirements build on existing Privacy Act transparency obligations rather than introducing an entirely separate regime.
Consultation, transparency, and accountability are expected to be the main focus for workplace uses specifically, rather than an outright ban on any AI use in these areas.
Why it matters for you: if you use AI tools anywhere in hiring, shift scheduling, or staff performance tracking, it's worth finding out now whether you'll need to disclose that to the people affected, rather than scrambling in December.
Source: Landers
🏛️ The Federal Government Has Outlined Plans for a Legislated "Digital Duty of Care" for AI Companies
The Daily Rundown: Announced on 20 July, the package puts the onus on AI companies themselves to build in safety by design, led by the Minister for Communications, alongside a second tranche of privacy law reform led by the Attorney-General and new attention on risks like retail surveillance pricing and AI-driven "agentic commerce." It's part of the broader National AI Plan the government has been rolling out through the year, following on from the AI Standards announcement earlier in July.
The Digital Duty of Care would be legislated, not just guidance, meaning it would carry legal weight once passed.
Treasury has previously said existing Australian Consumer Law can generally handle AI products and services, but flagged surveillance pricing and agentic commerce as specific gaps needing closer attention.
Why it matters for you: most of the direct obligations here land on AI providers rather than the businesses using their tools, but the surveillance-pricing and agentic-commerce mentions are worth watching if your business uses AI for pricing or automated purchasing decisions.
Source: Telecompaper, Landers
TOP STORIES
🌐 More Than 1,100 AI Employees Ask Governments to Help Pace AI Development, and OpenAI and Anthropic Just Backed Them
Demonstrators participate in a "Stop the AI Race" march in San Francisco on July 11. (Karl Mondon/AFP/Getty Images)
The Daily Rundown: Over 1,100 employees at OpenAI, Anthropic, Google, and Meta, including senior figures like Anthropic's co-founders and OpenAI's chief scientist, signed an open letter on 28 July asking the US government to help build the tools needed to deliberately slow AI development if it ever outpaces society's ability to oversee it safely. Both OpenAI and Anthropic then endorsed the letter as companies, not just through individual employees.
The letter doesn't call for an immediate pause, it asks for the technical and governance groundwork that would make a coordinated slowdown possible later if needed.
It follows closely on an incident where an OpenAI model reportedly breached another company's systems on its own, which several signatories cited as the kind of risk they're worried about.
Practical takeaway: you don't need to act on this directly, but it's a signal that the companies building these tools are taking the "moving too fast" concern seriously themselves, worth keeping in mind next time you see AI safety dismissed as pure marketing.
Source: Washington Post, Unite.AI
⚖️ Anthropic Pays $1.5 Billion in the Biggest AI Copyright Settlement Yet, Then Faces Backlash Over How the Books Were Sourced
The Claude Mythos logo is displayed on the screen of a smartphone placed on a reflective surface onto which the Anthropic logo. (Photo by Samuel Boivin/NurPhoto via Getty Images)
The Daily Rundown: Anthropic (the company behind the Claude AI models) agreed to pay $1.5 billion to settle a class-action lawsuit over using hundreds of thousands of pirated books to train its AI, the largest copyright settlement in US history. Days later, fresh backlash emerged after it came out that the company had also purchased rare physical books in bulk, scanned them with high-speed machines that sliced off their spines, then discarded the originals, a legal but heavily criticised practice.
A judge had found Anthropic's book-scanning process itself "transformative" and covered by fair use, the settlement was specifically about the pirated copies.
Authors, publishers, and artists have filed dozens of similar copyright lawsuits against AI companies over the past three years.
Practical takeaway: if your business uses AI-generated text, images, or other content commercially, this is a good reminder that the copyright status of AI training data is still being fought out in court, worth keeping half an eye on rather than assuming it's settled law.
Source: Stocktwits
🕵️ An AI Security Researcher Found Critical Flaws in Bing Before Any Human Did
The Daily Rundown: XBOW, an autonomous AI security research company, discovered three critical vulnerabilities in Microsoft's Bing search infrastructure, including two rated at the maximum severity score that could have let an attacker run commands on Microsoft's own servers with no login required. Microsoft quietly patched all three back in March, and XBOW published the technical details in late July once the fixes had rolled out fleet-wide.
The bugs came from a common image-processing shortcut, not something exotic, meaning similar flaws could exist in other businesses' image-upload features too.
XBOW is reportedly the first autonomous AI system to land in the top 10 of Microsoft's bug bounty leader board.
Practical takeaway: if your website or app lets customers upload images, it's worth asking whoever built it whether that upload path has been security-tested recently, this exact bug class has caught major companies before.
Source: The Hacker News, CyberSecurityNews
AI TOOL SPOTLIGHT
Perplexity
Perplexity, an AI search tool that answers your question directly in plain sentences, with links to where each fact came from, instead of handing you ten blue links to sort through yourself.
That makes it a handy alternative to a normal search engine when you want a quick, referenced answer to something specific, like a regulation, a statistic, or how a competitor does something.
It's particularly useful for small business owners who need to fact-check something fast without falling down a research rabbit hole. There's a solid free tier, with a paid plan for heavier daily use or more advanced models. perplexity.ai
PROMPTING TIP OF THE DAY
With copyright and AI training data back in the headlines, it's worth getting AI to check content before you publish it, not just write it.
Example (social media copy):
Instead of saying: "Write me an Instagram caption for my new range of scented candles."
Try this: "I run a small candle business in Geelong. Write me an Instagram caption for my new range of scented candles, then check what you've written against well-known brand slogans or existing marketing lines and flag anything that sounds too close to something already out there, so I can tweak it before I post."
You'll get copy that's still ready to use, but with a built-in check instead of finding out later it echoes someone else's line.
FROM READING TO DOING
If today's employer transparency rules have you wondering whether your business is ready for December, that's exactly the kind of practical prep Vortex Academy's courses covers, plain-English guidance on using AI responsibly, not just effectively. Have a look at what's on offer
