It is natural to trust certain familiar voices, but in this age of AI and ever more sophisticated forms of impersonation, risks arise.

The following article delves into the murky world of AI-assisted impersonation, a form of fraud that challenges the security systems of banks, wealth managers, family offices, and others. The article is from Dr Kenneth Dunkwu (pictured below), who is an entrepreneur, educator and professional voice artist. (His last area of expertise is particularly apposite for this commentary.) 

The editors are pleased to share these ideas and hope they stimulate conversations. The ways in which AI can be used for harm and also, hopefully, for good are important themes. The usual editorial disclaimers apply to views of guest writers. To comment, email tom.burroughes@wealthbriefing.com and amanda.cheesley@clearviewpublishing.com

Dr Kenneth Dunkwu

In September 2026, Reuters reported that fraudsters used AI-assisted impersonation in a scheme that led Fideuram, the private-banking arm of Intesa Sanpaolo, to transfer approximately €95 million ($106.5 million) overseas. Reuters reported that around €53 million had been recovered, while approximately €36 million remained missing. (1)

According to the report, the scheme began in February when the Fideuram chairman Paolo Molesini, at that time, received a WhatsApp message that appeared to come from Intesa Sanpaolo chief executive Carlo Messina, requesting urgent assistance with an overseas transaction. A subsequent telephone call appeared to come from a senior partner at a prominent law firm, whose voice had allegedly been replicated using artificial intelligence. Believing the request to be genuine, Molesini instructed the finance department to arrange the transfers.

The incident illustrates a growing risk for organisations whose decisions depend on trusted relationships and established authority. When a voice sounds familiar, the natural inclination may be to accept the person behind it as genuine. Yet a convincing voice can create an impression of identity without establishing it.

For family offices, where significant financial decisions often sit alongside personal relationships and sensitive family affairs, that distinction deserves particular attention.

The vulnerability of familiarity
Family offices operate at the intersection of wealth management, investment, property, legal affairs and family governance. Their work frequently involves coordinating decisions across jurisdictions, institutions and professional advisers. In smaller teams, responsibilities may be concentrated among a few trusted individuals, with relationships developed over many years.

That closeness can be a strength. Direct communication, discretion and an understanding of a family's priorities allow decisions to be made efficiently. However, the same familiarity can make an apparently authentic instruction particularly persuasive when it arrives unexpectedly or under pressure.

The information needed to construct a convincing impersonation may also be more accessible than assumed. Public interviews, conference appearances, recorded presentations, corporate announcements and professional communications can reveal voices, speaking styles, relationships and areas of responsibility. Details about business interests or professional networks can provide additional context that makes a fraudulent request seem plausible.

The potential consequences extend beyond a single transfer. An impersonated instruction might influence an investment decision, redirect funds, disclose confidential information or undermine confidence between a family, its office and its advisors. Even where a fraud is detected, uncertainty about which communications are genuine can damage relationships that depend on trust.

The challenge is not that family offices should abandon close relationships. It is that familiarity, while valuable, cannot independently establish the legitimacy of a consequential instruction.

When recognition is mistaken for authority
Three concepts are easily conflated in a high-pressure situation: recognition, verification and authorisation.

Recognition is the sense that a voice, name or manner of speaking is familiar. Verification establishes whether the person communicating is genuinely who they claim to be and whether the instruction is authentic. Authorisation determines whether that person is entitled to request the action in question.

These are related, but they are not interchangeable.

A voice may resemble that of a principal, family member or trusted adviser without proving identity. Even a genuine identity does not automatically establish that a particular transaction has been approved or falls within the person's authority.

This distinction becomes especially important where hierarchy and personal relationships influence behaviour. Employees may hesitate to question a request that appears to come from a senior principal, particularly when the instruction is framed as urgent, confidential or exceptional. The pressure to demonstrate responsiveness can displace the instinct to establish whether the request is legitimate.

The resulting vulnerability is organisational as much as technological. AI can help create the convincing voice, but the financial consequences arise when that imitation influences a decision before identity and authority have been adequately established.

Private wealth, public exposure
Privacy is central to family-office culture, but privacy should not be confused with invisibility. Family members, principals and advisors may have public profiles, attend industry events, give interviews or appear in recordings. Their businesses may publish announcements, and their professional relationships may be visible through public sources.

Individually, these details may appear unremarkable. Taken together, they can provide material for constructing communications that sound credible and reflect real circumstances.

Artificial intelligence changes the significance of recorded speech because a person's voice can potentially be reproduced without that person's direct participation in a conversation. The result may carry the emotional weight of a familiar relationship while offering no reliable proof that the individual is speaking.

This creates a particular tension for family offices. Their work depends on discretion and personal trust, yet those qualities can be exploited when a fraudulent communication appears to come from someone entitled to make decisions. Confidentiality remains important, but it cannot, on its own, establish authenticity.

The issue is therefore not simply how much information a family makes public. It is how readily a convincing communication can be mistaken for a legitimate instruction.

The governance implications
The discussion surrounding family-office security increasingly encompasses the relationship between technology, privacy and human judgement. A Family Wealth Report article published in October 2026, "Family Offices Must Have Privacy As Their North Star" – Summit Report, reflects contemporary concerns about the security implications of AI and the importance of protecting sensitive information. (2)

AI-enabled voice impersonation adds another dimension to that discussion. It brings the question of identity directly into the decision-making through which financial authority is exercised.

For family offices, this is not solely a cybersecurity issue. It concerns governance, accountability and the conditions under which an instruction becomes sufficiently credible to influence action. Technology may contribute to the deception, but the broader risk lies in the relationship between a persuasive communication and the human or organisational response it triggers.

The challenge is to preserve the advantages of trusted relationships without treating trust itself as proof. That distinction matters wherever people act on behalf of principals, manage substantial assets or handle decisions whose consequences may be difficult to reverse.

As AI makes convincing voice imitation increasingly accessible, the question facing family offices is no longer simply whether a voice sounds genuine. It is whether the identity and authority behind that voice have actually been established.

Protecting wealth in this environment depends not only on who appears to be speaking, but also on the integrity of the processes through which instructions acquire authority. The more convincingly a voice can be reproduced, the more important it becomes to understand what trust can establish and what it cannot.

References
1.    Reuters, “AI messaging scam costs Italy's top bank Intesa millions, sources say,” 25 September 2026. https://www.reuters.com/legal/government/ai-messaging-scam-costs-italys-top-bank-intesa-millions-sources-say-2026-09-25/
2.    Family Wealth Report, “Family Offices Must Have Privacy As Their North Star” – Summit Report, 21 October 2026. https://www.familywealthreport.com/article.php/%E2%80%9CFamily-Offices-Must-Have-Privacy-As-Their-North-Star%E2%80%9D-%E2%80%93-Summit-Report?id=209143

About the author
Dr Kenneth Dunkwu is an entrepreneur, educator and professional voice artist with more than 14 years of experience and credits spanning international brands including Amazon, Disney+, Mastercard, Armani and Emirates. He is the founder of VoiceShield AI, which focuses on AI voice impersonation risk and protecting identity, authority and consequential decision-making in high-trust organisations, including family offices and private wealth.