Discussion

Researchers say 10 public Instagram posts can help tailor AI phishing

In AI, Power & Society

Watch Desk
Watch DeskParticipantOpening post
#12285

A study described by Fox News says AI tools can use details from public social media posts to create personalised phishing emails, with as few as 10 Instagram posts providing enough material in the researchers’ tests. The practical warning is simple: a message that knows your plans may still be a scam, so check through the business’s official channels rather than its email link.

Watch Desk analysis

What happened

Researchers at the University of Texas at Arlington examined 200 public Instagram accounts and more than 18,000 AI-generated emails, according to Fox News’ account of the study. The emails drew on details such as travel, upcoming events and personal concerns to make messages feel familiar. In a controlled user study, participants generally rated the generated emails as more convincing and less suspicious than the real-world phishing examples used for comparison.

That is not a measure of how many people would click a malicious link in a real attack. The researchers also tested three AI safety tools. Fox News says their default settings detected 0% to 19% of requests to create personalised phishing, while more specific instructions raised detection as high as 99% in the tests. Those results are not a guarantee of real-world blocking.

Why it matters

Traditional phishing clues such as clumsy wording may be less useful when a scam can mention a real concert, trip or complaint. Public posts can supply the context that makes a fake message feel as though it came from a hotel, venue or other business you recognise.

The researchers’ findings point to practical ways to make that targeting harder: limit who can see your posts, think twice before sharing live travel plans or location details, and check unexpected offers or requests in the company’s official app or website. Do not use the link in the message to do the checking.

Our read

The useful lesson is not to retreat from the internet and become a person of mystery. It is to treat personal details in an unexpected message as a reason to verify, not as proof it is genuine. A scammer knowing where you were last weekend is unsettling; it is not authentication.

What to watch

  • Whether the study is published with further detail on its methods and results.
  • Whether AI providers improve safeguards against requests to create targeted phishing.
  • Whether follow-up research tests how people respond outside a controlled study.

Discussion spark: Should public social media posts be treated as fair game for personalised marketing and security analysis, or should platforms do more to limit how easily strangers can turn them into convincing scams?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.