The voice on the phone was not the boss, and the funds were all called the same day

Bloomberg reported on 5 August 2026 that several of Wall Street's largest funds faced attempted attacks through voice phishing, where AI-generated voices impersonate known people over the phone. Two Sigma, Citadel and Point72 are named in the coverage, and InvestmentNews reports that Millennium Management was among the targets as well. Several private equity firms are said to have faced the same. Two Sigma states that its security team responded quickly and that there is no indication that data or systems were affected, while Point72 has according to InvestmentNews informed its investors and states that an initial review found no client information stolen. Citadel and Millennium declined to comment.

What happens technically

Voice phishing is not new. What is new is the cost. Impersonating a specific person previously required either acting or technical craft, and both limited how many calls a single attacker could make. With synthetic voices that limit falls away. Voice cloning generally builds on audio recordings that already exist publicly, for instance an interview, a webinar or a podcast. That is an ordinary description of the technique, not a claim about how these particular calls were produced. It nonetheless explains the pattern in this case. Several funds were contacted on the same day, rather than one target being worked on over weeks.

The target in such campaigns is rarely a technical vulnerability. The target is a person with the authority to make an exception. According to the coverage the attackers tried to get employees to hand over information or access. The typical sequence in this attack type is that an employee is asked to read out a one-time code, reset a password or approve a sign-in already under way. Two-factor authentication is then no longer an obstacle, because the second factor is handed over willingly by the person it was meant to protect. That last part is a general description of the attack type. The sources do not say what was specifically attempted at each individual fund.

What actually happened remains partly unanswered. Two Sigma states the attempt was stopped. Point72 states an initial review found no client information stolen, and that the matter is still under review. Citadel and Millennium have not commented. None of the sources reports a confirmed breach. Absence of confirmation is not the same as absence of an incident, and the two should be kept apart. FINRA, the regulator for broker-dealers in the United States, contacted member firms about the attempts according to InvestmentNews. FINRA opened a centre for sharing threat intelligence among its members in March 2026, and this case is among the first occasions it has been used.

What this means for you if your employees answer the phone

The funds named here are among the best defended organisations that exist. They have budget, people and tooling far beyond what you command. Your attack surface is nonetheless the same as theirs, because it runs through a telephone and a person who wants to help. That is what makes this case relevant well outside finance. If you have a service desk, an accounts department or an on-call number, you have exactly the same surface. Berigo has previously covered how phishing toolkits are extended to get past two-factor authentication. We have also covered the arrests connected to the Tycoon2FA platform. The direction is clear enough. Attacks are moving from the technology to the process around the technology.

As we see it, the defence here is organisational before it is technical. A voice can no longer serve as proof of who is calling, and that should be written down somewhere everyone in your organisation can find it. What works is moving the confirmation into a channel the attacker does not control. Call back on a number from the directory, not on the number that called. Agree a fixed channel for urgent requests from management, so that a departure from the channel is itself the warning. Most important of all is that you remove the penalty for saying no. Most successful attacks of this kind succeed because the employee did not want to be difficult with someone who sounded like a manager. Under NIS2 article 21 this belongs to training and incident handling, but the real control is the culture that makes it safe to hang up and call back.

Berigo recommends

• Introduce a standing rule of calling back on a known number for any request involving access, codes or payments.
• Remove the ability for the service desk to reset strong authentication on the basis of a phone call alone.
• Agree a fixed channel for urgent requests from management, so that departure from it raises suspicion by itself.
• Rehearse the scenario of a voice that sounds entirely right, and let employees experience that hanging up is allowed.
• Log and alert when someone requests an exception to the routine, so repeated attempts become visible across departments.