Extract from ABC News
ASIC says it is taking down hundreds of phishing and investment scam sites every week, many of which use AI-generated content. (Supplied: ASIC)
In short:
Australia's financial regulator ASIC says it took down almost 12,000 phishing and investment sites last year.
Experts warn AI has turned cybercrime into a "supermarket aisle" where scammers pick and choose their attack.
What's next?
Regulators and offenders will continue to play a constant game of cat and mouse as the technology rapidly evolves.
Almost 12,000 scam websites have been shut down by Australia's financial regulator, as it warns artificial intelligence is "supercharging" online threats.
The Australian Securities and Investments Commission (ASIC) said it coordinated the removal of 11,964 phishing and investment scam websites in 2025.
The record number marked a 90 per cent increase in take-downs compared with 2024, when 6,270 sites were removed.
It averaged 32 websites per day, or 230 each week.
ASIC commissioner Alan Kirkland told the ABC the organisation had also seen a decline in the amount of losses reported by Australians due to investment scams of "around 11 per cent".
"So that says to us, [we have] to do more to stay on top of this, but we are actually having an impact," he said.
"We have a third-party provider that actively scans the web looking for investment scam websites.
"When it identifies them, it refers them to us, we check them, we confirm that in our view that [it is] a scam website, and then we instruct them to take it down.
"We do also receive reports from members of the public and from financial institutions like banks, and they get fed through the same process.
"It's really important for us to know that we're on the lookout for these websites."
Since launching that take-down service in 2023, ASIC has been forced to continuously evolve its approach.
Share this message to alert others to🛑STOP.✅ CHECK.🛡️ PROTECT. if they receive a call or message from the ASD’s Australian Cyber Security Centre. https://t.co/WRlhBwCTgO
— National Anti-Scam Centre (@nascgovau) September 10, 2025
AI, Mr Kirkland said, had begun playing two key roles in scams.
"One is that AI makes it much easier to create a new website," he said.
"But the other way is that AI is being used as a kind of claim to encourage people to make investments.
"We frequently see scam websites that spruik trading technology, where trading bots supposedly are able to generate returns that are higher and much faster than humans.
"Scammers are using the gloss of AI in order to sell something."
The AI 'supermarket aisle' for online scammers
ASIC says it has removed more than 25,000 investment scam and phishing websites since its take-down service launched in 2023.
It said it also took down more than 1,100 online investment scam ads on social media.
Australians lost a total $2.18 billion to scams in 2025, according to the National Anti-Scam Centre's latest Targeting Scams Report.
In that same time period, they reported 481,523 scams to various watchdogs.
That same report noted an increase in the "sophistication" of scams as a result of scammers using AI to generate their content.
Federal government watchdog Scamwatch has also warned of scammers using AI to generate fake videos, audios, and better tailor their scams.
Paul Haskell-Dowland is a professor of Cyber Security Practice and Associate Dean of International and Strategic Partnerships at Edith Cowan University.
"[Using AI] you can spin up a website, 10 websites, 100 websites, almost unlimited numbers, pretty much at the flick of a switch,"he said.
"Going back a few years … it required a lot of effort and invariably was done either manually or with cheap tools.
"They weren't very good. You could spot them a mile off, the messages didn't look convincing, the content wasn't very worthwhile.
"And some people would always be fooled by these things, but the vast majority would look at it and walk on by.
"The age of AI has completely changed that."
Creating scams has become a "service industry", according to Professor Haskell-Dowland.
"The tools that are out there have become very much a menu you can choose from," he said.
"It is very much like working your way down a supermarket aisle and just picking and choosing the pieces that you want, then letting the tools assemble it."
"It's actually a little worse than that because it isn't just that you can generate pretty much anything, it's that you can now optimise it for your target audience.
"Usually, what we're talking about here is large language models, the generative AI that we've now seen with things like ChatGPT and Copilot.
"Not only are you using the technology to generate the content specific to the market you're trying to attack … you could even utilise targeted attacks against individuals.
"If there were high-net-worth individuals you wanted to target for your financial scam, you can actually generate content that is optimised to them.
"[That's by] gathering information through LinkedIn, social media, public media, and then creating stories or products that you think would appeal to that individual."
'Cat-and-mouse game' between cybercriminals and online defenders
New laws passed last February would make tech giants, banks and other institutions liable to repay victims of scams.
Under those laws, dubbed the Scam Protection Framework, social media companies would be required to verify advertisers, banks must confirm the identity of payees, and telcos must detect and stop scam texts and calls.
However, it remains unknown when mandatory codes for each sector will be put in place.
That protection framework, according to Mr Kirkland, will be key to the changing fight against online scams.
"We expect the nature of scam threats to continue to evolve," he said.
"So that means as regulators we've got to be constantly on our game with how we're identifying scams and disrupting them and having websites taken down.
"The new obligations [from the scam protection framework] will be particularly important, because a lot of the scams that trick people start off with an ad on social media."
The technology, meanwhile, will keep progressing "rapidly", according to Professor Haskell-Dowland.
"When [ChatGPT] hit mainstream, that's when the technologies became available," he said.
"You look at any other technology, and moving to effectively free-to-access usually takes a very, very long time.
"With AI we've seen it almost appear overnight.
"I would say it's only been the last three to five years that we've started to see that being used by criminals to any great extent.
"The technology isn't going to slow down, our adoption [of it] isn't going to slow down.
"You think about any commercial product now that you go out to buy … you'd be looking hard to find something that doesn't have AI embedded within it."
Cybercriminals and those trying to fend them off, he said, were in a "cat-and-mouse game".
"There's always a level of one-upmanship," he said.
"Sometimes it's the cybercriminals who are one step ahead, and sometimes it is the cyberdefenders … but it will be a constant game.
"I wouldn't underestimate the potential [of a permanent solution] because no one saw AI coming, so it's quite possible that other technologies could come along that would either prevent it or potentially make it worse.
"I think it's a 'watch this space' situation, just see how things evolve over the coming years."
No comments:
Post a Comment