Institutional
A coordinated AI voice-cloning campaign hit Citadel, Point72, Two Sigma and Millennium Management within days of each other on August 5, forcing the first live test of FINRA's five-month-old Financial Intelligence Fusion Center
Attackers used AI to clone the voice, tone and phrasing of trusted executives in phone calls and audio messages, trying to talk employees into handing over credentials or network access. Two Sigma says it blocked the attempt outright, Point72 found no evidence of stolen client data so far, and Citadel and Millennium are not saying anything at all.
Four of the industry's largest multi-strategy and quant shops were targeted in the same window, alongside several unnamed private equity firms, which is the detail that turns this from an isolated incident into an industry event. The attack vector was vishing, voice phishing, upgraded with AI voice cloning good enough to replicate a specific executive's tone and phrasing on a live call or a voicemail, aimed at employees who would reasonably trust that voice enough to act on it. Align Managed Services president Vinod Paul's framing of the shift is the operative point, AI has moved this style of attack from something a criminal crew could run against a handful of targets to something that scales to thousands, because the expensive part, convincingly sounding like a specific person, is now largely automated.
The firms' responses split cleanly and that split is itself informative. Two Sigma says its security team caught the attempt and confirms no impact to data or systems, which reads as a team that had voice-spoofing specifically in its threat model already. Point72 disclosed the attack to investors and reported no evidence of stolen client information from its initial review, while acknowledging the investigation is ongoing, a disclosure posture that gives limited partners something concrete rather than silence. Citadel and Millennium declined to comment on whether their defences held at all, which is a legitimate operational security stance but also means the outside world has no way to independently gauge how close either firm came to a real breach.
FINRA's Financial Intelligence Fusion Center, a secure portal launched in March 2026 specifically so member firms could share fraud intelligence and coordinate responses, got its first real-world exercise here rather than a tabletop one. Whether the Fusion Center actually changed the outcome for any of the four firms is not something Fortune's reporting can answer given how little Citadel and Millennium are saying, but the fact that a coordinated attack landed within days of the Fusion Center's design brief being tested is the kind of validation regulators usually only get after something goes badly wrong. This time the wrong outcome, at least on current evidence, seems to have been avoided or contained.
The read-through for trading operations generally is that voice, not code, is now a credible attack surface for the firms that run the most consequential order flow in the market, and that credential and network-access controls built around email and system logins do not automatically extend to a phone call that sounds exactly like the CTO. Firms that have not yet built explicit verification protocols for voice-initiated access requests, a callback procedure, a shared passphrase, anything that does not rely on the caller merely sounding right, are the ones this campaign was implicitly testing. The ESAs' warning about frontier AI compressing the gap between vulnerability and exploit was published days before this attack; this is what that compression looks like when it reaches the trading floor rather than the server room.
Read the original: Fortune - A coordinated AI voice-cloning campaign hit Citadel, Point72, Two Sigma and Millennium Management within days of each other on August 5, forcing the first live test of FINRA's five-month-old Financial Intelligence Fusion Center. Commentary is the independent editorial view of Share Trading; the original article is credited to its publisher.