The Most Dangerous Financial Adviser You Have Never Heard of Is Already on Your Phone

AI chatbots are answering questions about retirement, debt, and insurance without authorisation, accountability, or any obligation to act in your interest.

Audio summary: https://youtu.be/SLCRvcjlWxw

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://open.substack.com/pub/johanosteyn/p/the-most-dangerous-financial-adviser?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

Picture an ordinary evening in a household somewhere in Johannesburg. A young professional, trying to work out whether to put extra money into a retirement annuity or pay down debt, opens a chatbot on their phone and types the question. The response arrives in seconds. It is confident, clearly explained, personalised in tone, and entirely unregulated. Nobody licensed that advice. Nobody can be held accountable if it turns out to be wrong for that person’s specific circumstances. There is no complaints process, no ombudsman, and no professional obligation to act in the client’s interest. The chatbot has no client. It has a user.

This is not a hypothetical scenario. It is happening now, at scale, and the regulatory frameworks designed to protect consumers from exactly this kind of harm were not built for it.

CONTEXT AND BACKGROUND

On 26 March 2026, the UK’s Financial Conduct Authority published its annual perimeter report, identifying the growing use of general-purpose AI for guidance on borrowing, saving, and investing as one of the key new issues falling outside its regulatory remit and posing risks to consumers. The FCA explicitly flagged that current perimeter boundaries may no longer be fit for purpose if these unregulated services begin to cause consumer harm, and called on the government to consider updating those boundaries accordingly.

Separately, the FCA launched the Mills Review in January 2026, a long-term examination of how AI could reshape retail financial services by 2030, asking specifically whether AI systems are providing services functionally equivalent to regulated activities such as advice or intermediation while remaining entirely outside the regulatory perimeter.

The Digital Regulation Cooperation Forum, bringing together the FCA, the ICO, the CMA, and Ofcom, has published early thinking on agentic AI, warning that AI-enabled financial services could capture significant value from regulated firms while remaining outside the perimeter entirely. The message from regulators is becoming clearer: the gap between what AI is doing and what the regulatory framework covers is widening, and the consumers in that gap have no meaningful protection.

In South Africa, the picture is no less urgent. The FSCA and Prudential Authority published their landmark joint report on artificial intelligence in the South African financial sector in November 2025, drawing on over 2,100 survey responses across banking, insurance, investments, payments, and lending. The findings were instructive: banks are leading AI adoption at 52 per cent, but governance frameworks remain deeply uneven, with 20 per cent of fintech firms reporting no AI governance at all, and only 5 per cent reporting mature oversight frameworks. The FSCA’s senior specialist on fintech noted that a model can be technically accurate while remaining impossible to explain to the consumer whose outcome it determined — a gap that is not merely a governance problem but a consumer protection failure operating in plain sight.

INSIGHT AND ANALYSIS

The regulatory framework for financial advice in South Africa was built around a fundamental premise: that consequential financial guidance should come from an authorised, accountable human being with a legal obligation to act in the client’s best interest. The Financial Advisory and Intermediary Services Act gives effect to that premise. It requires that financial advice be given by a representative of a licensed financial services provider, that the advice be appropriate to the client’s circumstances, and that there be a documented record capable of being examined, challenged, and if necessary, the subject of a complaint or claim.

AI financial chatbots satisfy none of those requirements. They are not licensed. They have no obligation of suitability. They do not know the user’s full financial circumstances unless the user volunteers them, and even then they have no framework for assessing what that information means for a specific individual’s situation. They cannot be held to account by the FSCA. They cannot be taken to the Ombud for Financial Services Providers. When the guidance turns out to be unsuitable and the user suffers financial harm, the architecture of redress that FAIS provides simply does not apply.

What makes this particularly significant in the South African context is not the availability of these tools to the wealthy and financially literate, who are most likely to understand their limitations. It is their availability to precisely the consumers who have historically been excluded from regulated financial advice entirely. South Africa has long had a structural problem with the accessibility of quality advice. Licensed financial advisers serve the middle and upper market. Large portions of the population have never had access to regulated advice on retirement savings, debt management, or insurance at all. AI chatbots are reaching those consumers now, filling a genuine and long-standing gap, while simultaneously stripping away every protection that made the regulated advice framework worth having in the first place.

The FSCA has signalled its approach: technology-neutral, outcomes-focused, designed to regulate better rather than more. That is a sound philosophical starting point. The challenge is that an outcomes-focused approach requires the ability to assess outcomes, and where AI operates entirely outside the regulated perimeter, the FSCA has limited visibility of what outcomes are being produced, for whom, and at what cost.

IMPLICATIONS

The liability question is no longer abstract. Air Canada was ordered by a tribunal to honour a discount its chatbot erroneously promised a customer. Tesla has been held liable for autonomous vehicle failures. A California jury in March 2026 held Meta and YouTube liable for algorithmic harm, awarding three million US dollars in damages. Closer to home, the FSCA has already raised alarms about deepfake videos of prominent figures endorsing fraudulent investment schemes, a pattern directly linked to the final liquidation of at least one South African financial services provider. The legal and regulatory environment around AI-related consumer harm is hardening globally, and South African financial institutions and technology firms operating in this space would be mistaken to assume that the current absence of specific AI legislation means an absence of exposure.

For boards and executives in financial services, the strategic question is not whether to use AI in customer-facing contexts. The technology offers genuine capability to reach underserved markets, reduce costs, and improve access. The question is whether the governance architecture around those tools — the explainability, the suitability assessment, the audit trail, the escalation path to a human adviser — is robust enough to withstand the scrutiny that is coming. The FSCA has indicated that firms must be able to explain how AI-driven decisions about consumers were made. That requirement will not become easier to satisfy as the tools become more capable and more autonomous.

CLOSING TAKEAWAY

Regulated financial advice exists because unregulated financial guidance causes harm. The advice gap — the space between the consumers who need quality financial guidance and the advisers who can afford to serve them — is real, persistent, and genuinely damaging to household financial resilience in South Africa. AI has the potential to narrow that gap in ways that would represent a significant social and economic benefit. The path to that outcome, however, runs through governance, not around it.

The chatbot on your phone does not know whether you can afford to lose the money it is suggesting you invest. It does not know whether the debt it is advising you to consolidate carries terms that will cost you more in the long run. It does not know your risk profile, your dependants, or your employment security. What it knows is what you have told it, and what it has been trained to say in response. That is not financial advice. It is financial-sounding content. Until regulators, platforms, and financial institutions collectively build the framework to distinguish between the two, the most dangerous financial adviser many South Africans will ever encounter is already in their pocket.

Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net