The hidden threat of AI to your money

By Malissa Conlin*

Yet another shock rippled through the AI industry early in September when a software engineer from AI company Anthropic quit, claiming that the technology could wipe out humanity in the next 10 years. 

As headlines about AI going rogue grow more frequent, they underscore the need for greater caution around the technology. Yet despite these risks, the convenience and easy access to knowledge it offers is very difficult to ignore.

So, I’m not here to tell you not to use AI or to stop using AI, but rather to touch on a hidden security gap that affects you whether you use AI or not. That’s because more of the companies that you’re dealing with are turning to AI to handle processes once handled by back-office.

At Brenthurst we do exactly that for certain non-critical, process-driven tasks, which are governed by tight rules.

The bottom line is that your exposure to AI risks are real, especially when vital personal information is fed into a company’s workflows powered by AI engines and agents.

What a responsible provider does

There are a few basic steps that all companies, especially in financial services, must have in place to ensure that your information is kept secure.

Firstly, a responsible financial services provider prescribes and controls the AI tools that staff are allowed to use. This is a simple IT governance principle that applies to all software that companies use when dealing with sensitive customer data.

Your name on its own is low risk. Add your account number or portfolio details, and the risk increases. A responsible provider treats that combination with real caution - not as business as usual.

So, no-one at a provider with strict AI protocols should be moving your documents onto a personal laptop or a personal AI account to get something done faster. If a task calls for AI, it must happen on tools the business has approved and controls, not their own account.

None of this is optional for a South African provider. The Protection of Personal Information Act gives you real rights over your own information. You are allowed to know what's collected and why, to see it and have it corrected, to object to how it's used. Under Section 71, which deals with automated decision-making, you can even object to a decision that's been made entirely by a computer with no person involved.

At Brenthurst, we use approved AI tools with rules in place not to share any information with the AI provider. These tools are mainly used for checking policy wording and grammar, and to support training and non-confidential compliance content.

FICA records, banking details, complaints, and legal advice never go anywhere near it.

Brenthurst also has not and will never use AI to provide financial advice or make financial recommendations. AI may be a useful tool, but it should not replace professional advice or the judgement of a qualified adviser.

What to do yourself

I always urge clients to exercise caution about what information they hand over, especially in their own AI chats. Your ID number, banking details and full financial statements should never be shared, no matter how convenient it might seem at the time.

If you want a document summarised or a number checked, ask someone who's actually accountable for getting it right.

Treat a confident-sounding answer from AI as exactly that - confident-sounding, not necessarily correct. It doesn't know your full financial picture, your history, or what you're actually trying to achieve, and it faces no consequences if it gets your situation wrong.

A simple test covers most of it: General questions and drafting are fine to hand to AI. But when it comes to your actual numbers, or a decision about your money, the human element still matters – it goes to a person instead.

It is also very important to be aware of potential consequences of entering private, confidential, or proprietary information into AI tools. Investors should be particularly careful not to upload personal financial information, legal documents, Wills, trust documents or any intellectual property belonging to themselves, their family, or another person/entity.

Once confidential information is entered into an external AI platform, there may be risks around privacy, confidentiality and how that information is stored or processed. In the case of someone else’s information or intellectual property, there could also be legal and confidentiality implications for sharing it without the necessary permission.

None of this means avoiding AI, and it doesn't mean assuming the worst of the people managing your money. It means knowing what good practice looks like, so you can recognise it, whether it's in a chatbot's answer or in the provider you trust with your future.

* Malissa Conlin is a Director and the General Manager of Brenthurst Wealth, a qualified attorney, and Fiduciary Practitioner of South Africa® (a registered member of FISA) malissa@brenthurstwealth.co.za