Don’t blindly trust AI recommendations. Scammers are already poisoning the system
Criminals are mounting ‘poisoning attacks’ on AI systems where they trick them into linking to websites that look like the real thing, but are simply there to steal people’s money, writes head of the National Trading Standards Scams Team Louise Baxter
Late last year, National Trading Standards carried out research that found more than two-thirds of UK adults have been targeted by a scam or fraud, and more than one in five have had money stolen – rising to one in three under-34s.
Listen to this article
With fraud now the most common crime in the UK, society was already facing an enormous challenge. And that’s without the oncoming onslaught of AI-powered scams.
We are increasingly seeing criminals using AI to develop ever more sophisticated scams, such as building fake websites in minutes or creating fraudulent direct debits by cloning victims’ voices.
So when we were approached by the scam-checking platform Ask Silver with evidence that major AI models such as ChatGPT are being ‘poisoned’ to recommend scam websites, we wanted to ensure consumers are aware of this emerging threat and, most importantly, can take steps to protect themselves.
People are understandably enjoying the convenience of using AI to find answers to questions or get product recommendations quickly. But criminals are exploiting this by mounting ‘poisoning attacks’ on AI systems where they trick them into giving wrong or unsafe answers - such as linking to websites that look like the real thing, but are simply there to steal people’s money. Research suggests this ‘data-poisoning’ can be done with only a small amount of harmful content, making it a growing risk for both consumers and honest businesses.
We are urging anyone using AI for online searches to check that any recommended website is genuine before buying anything. The best defence is always to stop and think, and there are some simple steps you can take to help spot the fakes:
- Be suspicious of prices or deals that are far below what you’d expect
- Use a scam checker like Ask Silver or Get Safe Online
- Use review websites like Trustpilot or Google Reviews
- Check the domain name for misspellings, extra characters, or unusual extensions
- Check the site for bad spelling, grammar errors, low-quality images, or awkward layouts
- See if you can clearly find details for the registered company
- Check if they have clear shipping, privacy and returns policies
- Check the checkout process - legitimate sites offer secure payment options like credit cards or PayPal
- For scam alerts, advice and training on your phone, download the National Trading Standards’ Friends Against Scams app.
- VERIFY VERIFY VERIFY
As AI makes fraud and scams ever more sophisticated - and much harder to spot - anyone can become a victim. That's why we must put an end to victim-blaming language that leaves people feeling ashamed and puts them off reporting, and provide better fraud education in schools and for adults to build the nation’s long-term resilience against scams.
____________________
Louise Baxter MBE is Head of the National Trading Standards (NTS) Scams Team and the Lead Officer for Consumer Education, Advice Scams and Consumer Vulnerability at the Chartered Trading Standards Institute (CTSI).
LBC Opinion provides a platform for diverse opinions on current affairs and matters of public interest.
The views expressed are those of the authors and do not necessarily reflect the official LBC position.
To contact us email opinion@lbc.co.uk