AI may have become the most accessible thinking partner most of us have ever had, but that same technology is now in the hands of the people trying to defraud you.
Kurt Dodgen, Senior: Business Information Security Officer and Brian Abdoll, Group Information Security Programme Manager at Sanlam unpack the risks of using AI. Both are clear, the message is not to avoid AI, it’s to clearly understand what you are handing over, and slow down before you act.
For years, security awareness training taught people to look for the obvious tells, such as clumsy grammar, odd typos, a domain name that was almost-but-not-quite right, or even corporate colours slightly off in a spoofed brand.
"Now with the rise of AI, that's difficult to even detect, because the AIs are producing this phishing content," says Dodgen. “Generative tools now write fluent, convincing messages at scale. The checklist most of us carry in our heads was built for a threat that no longer looks the way it used to,” he explains,
The changes have been coming thick and fast. Early concerns three years ago centred on realistic phishing at volume. Today the conversation has moved to agentic AI threats. Now AI systems are being used to break into IT systems and corporate infrastructure, with potential for serious operational damage, not just individuals being tricked.
AI has made executive impersonation cheap and convincing. Abdoll says customers now need to be on the lookout for cloned voices on a phone call; a familiar face endorsing an exclusive investment opportunity; and a WhatsApp group that looks like it is run by people you recognise. More recently the team says they have been battling info-stealer malware, which is software that secretly lifts saved credentials straight out of your browser, then sells or uses them for account takeover.
Scam | What it might look like |
Fake emergency calls | A caller uses an AI-cloned voice to sound like a loved one in trouble and asks you to send money urgently. |
Convincing phishing messages | An email, SMS or WhatsApp message appears to come from a trusted organisation and asks you to click a link or share personal information. AI can make the wording polished and personalised. |
Fake investment promotions | An AI-generated video appears to show a celebrity or public figure promoting an investment, making a fraudulent offer seem credible. |
Romance and friendship scams | A scammer uses AI-generated photos and messages to create a convincing online identity, build your trust and eventually ask for money or promote a fake investment. |
Impersonating a boss or colleague | A voice message or video call appears to come from someone at work, asking you to make an urgent payment or share confidential information. AI can help scammers imitate that person’s voice or appearance. |
Tax season is a good example of when people should be extra wary. Many people now upload bank statements, payslips, ID numbers and proof-of-address documents to public AI tools for help with a tax return or a budget.
“This is not about being paranoid, it is about knowing how the AI company handles your information. If you can’t be sure whether it uses your data to train its models or has strong cybersecurity practices, your personal and financial information could be at risk,” Dodgen warns.
Abdoll says users shouldn't be swayed by the speed and convenience of AI tools. "Users need to understand the value of the information they're handling. You need to apply a practical test and ask yourself if you mind if the information ends up in the public domain. If you don’t then that’s fine, but if you don’t want to share your bank details or your personal medical information with strangers, then don’t share it with ChatGPT or any other AI tool,” he says.
The cyber security specialists share four golden rules to apply when you are asked to click a link, open an app, approve a transaction, or release a payout:
1. Refuse the urgency. Attackers manufacture time pressure deliberately and use urgency as a means to compromise their victim. A message that insists you act right now should be a red flag.
2. Switch channels to verify. If a message arrives by WhatsApp, phone a listed company number to verify the message is real. If it arrived by email, use the secure in-app chat or the contact centre. Never verify using the contact details supplied in the suspicious message.
3. Ask whether you were expecting it. A confirmation relating to a home loan or life cover application you are actively working on is plausible. An out-of-the-blue "approve your investment withdrawal payout" message is not.
4. Use the relationships you already have. If you work with a financial adviser, they are a verification channel — make use of your existing relationship and your channel to trusted advice.
Even the most vigilant of us may fall victim to a cyber scam at some point. It’s important to stay calm and take action.
Who to call if you think your personal information has been compromised or you have been defrauded:
Your Bank: Call your specific bank's fraud or customer care line right away to freeze or secure accounts and cards.
Southern African Fraud Prevention Services (SAFPS): Call 011 867 2234 or email protection@safps.org.za to apply for free Protective Registration which alerts credit providers that your identity has been compromised
South African Revenue Service (SARS): If your tax or ID details are affected, call the SARS Contact Centre at 0800 00 7277 or the Fraud Hotline at 0800 00 2870
How to find a financial adviser who understands my needs You can visit this link to find someone near you.
If you want to find out more about how you should be protecting yourself against increasingly sophisticated cybersecurity risks, why not join the conversation with the experts on the next Sanlam Make it Make Cents webinar.
This is general information and education only and does not constitute financial advice. For personal recommendations, speak to a licensed financial adviser.
Agentic AI: An AI system designed to plan and carry out a sequence of tasks with a degree of independence.
Phishing: A fraudulent message or website designed to trick someone into sharing information, opening a harmful attachment or clicking an unsafe link.
Deepfake: AI-generated or altered audio, video or images that imitate a real person or create an event that did not happen.
Voice cloning: The use of AI to imitate a person’s voice from recordings or audio samples.
Spoofing: Disguising a message, website, telephone number or email address so that it appears to come from a trusted source.
Information-stealing malware: Malicious software that secretly collects passwords, browser data, financial information or other sensitive details from a device.
Account takeover: When a criminal gains access to someone else’s online account and uses it without permission.
AI hallucination: A confident-sounding but inaccurate, misleading or invented response produced by an AI tool.
Information Regulator South Africa, “What is personal information?” Available at https://inforegulator.org.za/popia/
Sanlam, “Phishing: Stay Safe Online”
https://www.sanlamonline.co.za/stay-safe-online/how-to-spot-an-unsafe-link
Sanlam, “Scams”
https://www.sanlamonline.co.za/stay-safe-online/scams
Sanlam, “Get Advice”
https://www.sanlamonline.co.za/personal/get-advice
Sanlam On24 Webinar “Sanlam Make it Make Cents webinar”