Your email knows where your money is.
AI could help hackers find it

New Delhi, Aug 10, 2026

Barracuda controlled attack shows how attackers can weaponize an AI assistant to turn a single compromised account into CEO fraud and a $250K wire transfer scam

A compromised email account can already expose your bank details, investment documents and financial conversations. Now imagine giving a hacker an AI assistant that can read the inbox, identify the people in charge, find the next big payment and draft an email in the account holder’s own writing style. That is no longer a theoretical risk.

A controlled attack simulation by cybersecurity firm Barracuda has demonstrated how an attacker who gains access to an AI-enabled email account could use the AI assistant to move through an organisation, identify senior executives, uncover sensitive financial information and ultimately redirect a legitimate payment.

In the simulation, the attacker used Microsoft's Copilot after compromising an employee's account. The AI assistant helped identify the company's organisational structure and senior executives, draft a convincing phishing email and eventually gain access to the CEO's account. Once inside the CEO's mailbox, the attacker used Copilot to search financial correspondence and surfaced a pending **$247,500 wire transfer** awaiting final approval. The payment was then redirected to an account controlled by the attacker.

For investors, business owners, high-net-worth individuals (HNIs) and family offices, the warning is particularly relevant. Email is no longer simply a communication tool. It can contain a detailed map of a person's financial life — investment statements, property transactions, invoices, tax documents, bank details and conversations with wealth managers, lawyers and accountants.

And AI can make that information much easier for a fraudster to exploit.

The scam starts with a legitimate email account

The worrying part of Barracuda's experiment is that the attacker did not need to create a fake CEO email address.

After compromising an employee's account, the attacker used Copilot to understand the company's internal structure and identify the CEO as a valuable target. The AI then helped draft a phishing email based on an existing conversation and the employee's natural writing style.

Because the message came from a genuine internal account and contained information relevant to an ongoing conversation, it was more difficult to distinguish from legitimate communication. The CEO clicked on the link, allowing the attacker to take over an authenticated session and gain access to the CEO's account.

The attacker then used another AI-generated inbox rule to hide login alerts, making it less likely that the account holder would realise the account had been compromised.

Then AI goes looking for the money

This is where the experiment becomes particularly relevant to wealth management.

Once inside the CEO's mailbox, the attacker did not have to manually search through years of emails. They asked the AI assistant to summarise recent financial emails, including invoices, amounts and upcoming transfers.

Within seconds, the AI surfaced sensitive financial information, including the pending **$247,500 payment**.

The attacker then used the compromised CEO account to send the finance team a message requesting that the bank account details for the payment be changed.

The email came from the CEO's genuine mailbox, referred to a real transaction and matched the executive's usual tone. The finance team therefore changed the beneficiary details and sent the payment to the attacker's account.

The attackers also created another forwarding rule so that replies from the finance team would be diverted away from the CEO. They could therefore continue the conversation and maintain the appearance of a legitimate transaction.

Barracuda's conclusion is significant: the AI assistant did not give the attacker new privileges. Instead, it made it much faster and easier to exploit the access the attacker had already obtained.

" A user’s email history is full of sensitive information and context that can be leveraged by attackers, including emails sent and received, documents shared, attachments, and calendar invites. Company structure can be deduced from implied relationships in messages or directly viewed via organizational charts,” said Daniel Avulov,Senior Cybersecurity Researcher, red team at Barracuda.

“The controlled attack shows how AI assistants can become unwitting malicious insiders and improve both the quality and speed of an attack. The most effective defensive approach is to recognize that attack patterns remain the same, take advantage of the telemetry that already exists, and ensure mean time to detect is as low as possible.”

Why this matters to HNIs and family offices

For an individual, an email takeover can mean stolen passwords, personal information or fraudulent payments.

For someone with substantial wealth, the potential exposure is much larger.

A single inbox could contain: 

•     Details of upcoming property purchases or sales

•     Investment and brokerage correspondence

•     Bank account and beneficiary details

•     Mutual fund and portfolio statements

•     Tax and estate-planning documents

•     Insurance policies

•     Loan and financing documents

•      Invoices and payment instructions

•     Communication with private bankers and wealth managers

•     Details of large upcoming transactions

For a family office, the information can extend across several members of a family and multiple businesses.

An attacker who gains access to an employee's mailbox could potentially identify which banker handles a particular account, which lawyer is working on a property transaction, who approves payments and when a large transfer is scheduled.

The inbox becomes an intelligence repository — and AI becomes the tool that helps the attacker make sense of it.

India is already seeing the 'boss scam' problem

In June, India's Indian Cyber Crime Coordination Centre (I4C) warned about a "boss scam", in which cybercriminals impersonate senior executives or regulators to persuade employees to make high-value financial transfers. The warning specifically highlighted the vulnerability of finance departments.

The difference with the Barracuda scenario is that the fraudster may not even need to impersonate the executive from scratch.

If the executive's actual email account is compromised, the attacker can potentially send the instruction from the genuine mailbox, use real information from previous conversations and reproduce the person's communication style.

That makes traditional checks such as looking at the sender's email address considerably less useful.

AI is also being used to fight financial fraud

The technology is becoming a double-edged sword.

In May, I4C and the Reserve Bank Innovation Hub (RBIH) signed an agreement to strengthen AI-driven detection of mule accounts and cyber-enabled financial fraud. The collaboration involves sharing intelligence and using data to improve fraud-risk assessment systems.

The development shows how both sides of the fraud battle are adopting AI — banks and regulators to detect suspicious activity, and criminals to make social engineering and impersonation more convincing.

CERT-In has separately warned users about AI-generated phishing content, fake websites and impersonation attempts. It advises people to be particularly careful with urgent requests involving financial transactions or sensitive information and to independently verify such requests.

What should investors do?

The lesson is simple: **don't treat an email as proof of identity when money is involved.**

1. Verify every change in bank details

If a vendor, employee, lawyer, investment manager or business partner asks for a change in beneficiary or bank-account details, verify it independently.

Call the person using a number you already have — not a number included in the email.

2. Use a two-person approval system

For large payments, one person should initiate the transaction and another should independently verify the beneficiary details.

This is particularly important for family offices and privately held businesses where a small number of people control substantial sums.

 3. Use the RBI's beneficiary-name verification facility

For RTGS and NEFT transactions, the RBI has introduced a facility allowing customers to check the beneficiary's bank-account name before making a transfer. Banks were required to make the facility available through internet and mobile banking by April 1, 2025.

This provides an additional check against sending money to the wrong account.

4. Turn on multi-factor authentication

MFA remains an important security layer. CERT-In recommends using strong, unique passwords and enabling MFA wherever it is available.

However, the Barracuda simulation also demonstrates why MFA should not be treated as an absolute safeguard: attackers used a session-token takeover to bypass MFA after the CEO clicked the phishing link.

5. Don't approve an 'urgent' payment on email alone

A request such as "change the account details immediately", "make the transfer today" or "I'm unavailable, please handle this urgently" should trigger an additional verification step.

The more urgent the request, the more important it is to slow down.

6. Don't keep your entire financial life in one inbox

For wealthy individuals and family offices, concentrating investment statements, property documents, tax records and banking correspondence in a single email account creates an attractive target.

Limiting access and separating sensitive financial workflows can reduce the damage if one account is compromised.

The new wealth-protection rule: verify the person, not the email

The biggest takeaway from Barracuda's experiment is that **a genuine email address is no longer necessarily proof that a genuine person is behind the message**.

If an attacker controls the account, the email can come from the correct address, contain accurate information and sound exactly like the account holder.

[The Business Standard]