Dr Karen Sutherland

24/08/2026

This morning Ahron Young and I chatted about two news headlines on Ticker.

Both came back to the same problem: we are building systems faster than we are building the accountability around them.

On AI agents spending your money:

You tell an agent to find something.
You tell it not to buy without your approval. It buys anyway.

The retailer sees a valid order. The payment service sees a valid charge.
The agent provider holds your instruction.
None of those records talk to each other.

Where it gets worse:

Agents are very good at finding blurry areas in your instructions and filling in the details themselves
Some websites carry hidden code designed to derail an agent and push it into making purchases
Returns policies were not written for this

"My AI agent did it" is not a returns policy.

On South Australia's royal commission into AI:

Peter Malinauskas went to Silicon Valley, photographed with Tim Cook at Apple Park, and came home and announced a royal commission.

Royal commissions do not happen overnight. The technology is moving rapidly. By the time the report lands, will it still describe the world we are in?

The education focus is worth watching. How students use AI without getting it to do the learning for them.

What I see in my own students:
The ones who struggle are not struggling because they used AI.
They did not engage with the material, so they could not prompt properly or tell right from wrong in the output.

You need to be the subject matter expert to use AI well.

Every government conversation right now is about what we can stop AI doing. Very little is about what we could harness.

Reducing the harm matters. So does building the capability.
We are only having one of those conversations.

University of the Sunshine Coast Dharana Digital

17/08/2026

Cloud walking.
🫶🧡✨💕

17/08/2026

This morning Ahron Young and I chatted about two news headlines on Ticker.

Both came back to the same thing: the technology keeps moving faster than the protections built around it.

On AI reading your holiday photos:

McAfee scanned 21,000 travel images across two models.
Not the Eiffel Tower. Just a beach somewhere or a random building.

One model identified the location 91% of the time. The other 87%.

It is not using metadata.
Strip the location tags off entirely and the AI still works it out from the image.

How the scam runs:
You post an innocent holiday snap
AI identifies where you are
You get a text saying your bank detected suspicious activity while you were travelling there
The specificity makes it feel legitimate

It is a magic trick. Establish that everything is normal, then take what you came for.

What helps:
-Share with friends only, and post after you get home
-Treat urgency as the giveaway
-Never use the link. Ring the number on the back of your card

On France blocking its under-15 ban:
The Constitutional Council knocked it back for two reasons.
It deemed the ban a limitation on freedom of speech for young people, which our government did not weigh the same way, and the law was not clear on how age would actually be verified.

They have not said no. They have said come back with more information.

What I like about the European framing: they are calling it a "social media delay" rather than a ban. A softer way to approach the same problem.

Reports suggest at least a third of Australian children under 16 are still on the platforms. Some with new accounts, pretending to be adults.

Even the ones who are off social media are still on Roblox, still chatting, still shouting at each other online.

Kids will find a way to communicate. That was always true.

A ban on its own does not educate young people, support parents to supervise, or give kids other ways to connect.

The theme running through both stories:
Where there is a will, there is a way.

Scammers are using the same technology built to stop them. Kids are getting around bans built to protect them.

Vigilance and education will carry more weight than any single rule.

University of the Sunshine Coast Dharana Digital

Want to create live streams like this? Check out StreamYard:

16/08/2026

Every Mon morning I open my uni inbox and face a dumpster fire. 🔥

My out-of-office goes on Friday at 5pm, and ends Mondays at 8am.

An email-free weekend has been my gift to myself for a few years.
✨PURE BLISS ✨.

However...there's always a however. 😂

It also means two days of student enquiries, international colleagues in different time zones and everything else, all sitting there waiting for me when I return.

I used to just start at the top and work down.

Now I ask AI first. What actually needs me today?

Sometimes I do it on my phone over coffee, before I'm even at my desk.

By the time I open the laptop I already know where to start.

That's only one small shift. In this month's free AI webinar I walked through four:

1️⃣ Triaging your inbox
2️⃣ Drafting with real context
3️⃣ Humanising the output so it sounds like you
4️⃣ Optimising subject lines and send times


Don't get me wrong, AI still needs you.

It gets things wrong and it makes things up, and when an email goes out, it's from YOU.

You must stay the last decision maker, ALWAYS.

Recording is live now. Watch it here:

This is NOT a webinar about using AI to write robotic emails that b...

14/08/2026

The light will always shine through.
🫶🧡✨💕

12/08/2026

Rogue Agents. Addictive Design. Two Industries Marking Their Own Homework.

This morning Ahron Young and I chatted about two news headlines on Ticker.

Both stories came back to the same question: who is actually checking?

On AI agents breaking out of their sandboxes:

OpenAI and Anthropic have both been called before United States lawmakers.

OpenAI was testing an autonomous agent in a sandbox. The agent broke out and went onto the internet to try and solve one of the tests it had been set.

Anthropic ran an assessment across roughly 141k+ tests and found three breaches where agents had done the same thing.

These two companies are the ones who told us.
There is no legislation requiring any of it.
OpenAI reported when they really did not have to.
We only know what we are told.

This has been the year of the autonomous agent.
They have been in development for a while. This is where they have hit their stride.

Remember when OpenClaw was released, and a Meta security expert posted that it had deleted all of her emails without her consent because it was trying to free up room?

The unpredictability is what has everyone concerned.

The real risk for businesses:
1. You give agents access to your emails
2. You give them your files
3. You give them your sensitive information

The upside is real. It gets to know you,
it becomes more relevant, it saves you time.

The trade-off is that you have handed sensitive information to something that can behave unpredictably.

That weighing up is the conversation most organisations have not had yet.

Governments want to stop AI causing harm. Meanwhile these companies have promised investors that AI will soon think like humans, so businesses will need fewer of them.

More testing at that frontier means more breaches. That is the direction of travel.

Closer to home:
South Australia has announced a royal commission into AI. Victoria will have a Minister for AI in Ben Carroll's new front bench.

The question I have is how any of them will know what is going on.

Industry and international standards exist. The legal side still has to catch up with the technology.

On the social media lawsuits:
More than 3,000 cases against Meta and other platforms have been allowed to proceed.

The critical detail is that they are not examining content. They are examining design.

How the platforms were built. How the algorithms surface content.
Why young people cannot put their phones down.

Where this could end up:
-Significant financial penalties
-Platforms forced to redesign with young people in mind
-Or more countries simply banning under-16s, which we already know is hard to enforce

The argument the platforms will run:
That the young people who struggled were struggling for other reasons too.

Ahron put it well. If there is a crack in the road and you fall in it, who is responsible? You for not seeing it, the council for not fixing it, or whoever made it?

Where does platform responsibility stop and parental supervision begin? That question will take a long time to answer in court.

Commentators who worked at Meta have said mental health experts emailed executives to flag concerns and were ignored. Allegedly.

If that evidence holds up, this stops being a debate about design philosophy and becomes a question about what the companies knew.

The theme running through both stories:
In each case, the people best placed to identify the harm were the ones building the product.

Transparency should not be optional.

University of the Sunshine Coast Dharana Digital

Want to create live streams like this? Check out StreamYard:

10/08/2026

🚨 CLIENTS NEEDED FOR UNI SOCIAL MEDIA STUDENTS. 20 PLACES LEFT. 🚨

Want to improve your social media performance?

University of the Sunshine Coast social media students can help.

Register your interest here: https://forms.gle/AtDXnPkDMZyBj2tt5

Most organisations track followers and likes because those numbers are easy to find.

Very few can connect their social media activity to a goal that actually matters to the organisation.

This is what I'm attempting to my students to address this trimester.

Registrations are open for organisations wanting to partner with students in CMN308 Social Media: Monitoring and Measurement at UniSC.

Your organisation can be based anywhere, as the work happens online.

How it works:

🔺 You nominate one goal you would like to achieve through social media.

🔺 You are matched with a student who develops a plan to measure your performance against that goal

🔺 Your student implements the plan, then gathers and analyses the data.

🔺 You receive a report with findings and practical recommendations to strengthen your social media performance.

What we need from you:
1️⃣ Regular communication with your assigned student throughout the project.
2️⃣ Some data, or 'Analyst' access to Meta Business Suite Insights if you are comfortable providing it.

Every student signs a confidentiality agreement developed by the University before the project begins.

Businesses, charities, community groups and sporting clubs are all welcome.

Only 20 places remain.

Register your interest here: https://forms.gle/AtDXnPkDMZyBj2tt5

Please answer the following questions to register your interest in becoming a client for students from the University of the Sunshine Coast undertaking the course CMN308 Social Media: Monitoring and Measurement. If your business, charity, community or sporting group is selected, you will be matched....

07/08/2026

Contentment is the key to abundance.
Sw. Sivananda.
🫶🧡✨💕

05/08/2026

Undefined AI. Unenforceable Bans. Regulation Is Losing.
This morning Ahron Young and I chatted about two news headlines on Ticker.

Both stories came back to the same problem: governments trying to control something that's moving faster than they are.

On the White House calling in the AI platforms:
Anthropic, OpenAI, Google and Meta are being asked to hand over their new frontier models for assessment up to 30 days before public release.

The idea is to test whether these models can find holes in security systems.
However, nobody has actually defined what a frontier model is.
They also haven't said what the tests involve. That's classified.

And, it's completely voluntary.

Why now?
OpenAI came out last week and said one of its agents got out of the sandbox and went into Hugging Face to try and beat a hacking task it had been set as a test.

Anthropic had a stoush with the US government over Fable 5, which was pulled from public use for a while.
I think they're a bit scared about what's actually going on.

They want to see if these models can find holes in their own security before the public gets free rein to do the same.

The bigger question Ahron raised:
When governments get involved in technology, they're usually behind. It can stifle innovation. Or it can turn government into the kingmaker, deciding who gets to build what.

Also, because it's voluntary, the interesting part will be what happens to the platforms who say no.

On children, social media and school results:
A study out of Italy followed 5,000 children.

Kids who started using social media at 11 or 12 were about six months behind in maths and in Italian reading and writing by the time they turned 16.

Interestingly, their English improved. A lot of social media content is in English.

The issue was distraction more so than social media content.

The constant checking of the phone, not being able to focus on a task long enough to go deep into it.

Can a ban fix this?
It's very difficult to enforce. Every platform is different, and they all use different measures.

Also, it's not only social media.
It's the phone.

If you take social media away, they're still texting. Still gaming. Still scrolling.

A holistic approach is missing.

If you're taking social media away, what are you teaching kids to do instead?
How are they going to communicate and build relationships?

What are you putting in its place?

Schools have been doing their bit for years.
In Queensland, phones have been out of classrooms for a while now.
The social media ban has just been added on top.

The theme running through both stories:
Regulation is arriving after the technology, not before it.
In both cases, nobody has properly defined what they're actually regulating.


University of the Sunshine Coast Dharana Digital

Want to create live streams like this? Check out StreamYard:

29/07/2026

Predators Are Deepfaking Students' School Photos. AI Doctors Are Selling Dodgy Products. The Horse Has Bolted.

This morning Ahron Young and I chatted about two news headlines on Ticker.

Both stories explored the impossibility of stopping Ai generated content once it's already out there.

On AI deepfakes targeting students:
Australia's eSafety Commissioner has received more than 100 reports in the last three months of people using school photos of children to generate deepfakes.

Schools have been put on alert to think very carefully about using photos and videos of real children online.

One school has already acted:
An educational body here in Queensland stopped using real photos of children for promotions like school uniforms.

They now use AI-generated images of children, and they're transparent about it. They disclose it.

They're fighting AI with AI.

Why this matters:
Earlier this year, a girls' sports team was on the podium celebrating a win.
The school received a blackmail threat. The girls were 10 or 11.
Someone threatened to use the photos in X-***ed material unless the school paid up.

This is very real. It's happening.

Can the law keep up?
Fines can reach $54.6 million, but this stuff is instant.
It keeps getting faster, more believable, easier to access.

The law is too slow.
There are so many tools out there, and people can build their own if they have the knowledge. You can't even document how many exist.

The horse has bolted. It's too hard now to regulate what's already out there.

On AI health influencers:
AI-generated doctors are misleading millions seeking medical advice.

They're not real. They're created to sell products. Usually dodgy ones.

The risk:
People believe outlandish claims because they think it's a real doctor.
They're losing money and putting their health at risk.

Real doctors face a credibility problem too.
They may not be believed when they post legitimate content, because people can't always tell the difference.

Some doctors are fighting back on platforms like TikTok, critiquing the AI doctors and the products they push. Showing people the difference.

Can AI influencers be held accountable?
Only if you can find who created them, and even then, they can delete the account and start a new one under a different username.

For anyone determined to keep doing this, finding them is very difficult.

The theme running through both stories:
We're talking about the inability of authorities, platforms, anyone, to combat misinformation online.

The only advice:
Check. Ask people you trust. Do the background work.
It may not be enough, but it's what we've got right now.

University of the Sunshine Coast Dharana Digital

Want to create live streams like this? Check out StreamYard: