AI impersonation fraud: recognizing a name or a voice is no longer verification

What happened

Two warnings show AI impersonation fraud targeting both officials and finance teams.

On December 19, 2025, the FBI issued a public service announcement (I-121925-PSA), updating its May 2025 warning about a campaign impersonating senior US officials. It says that “since at least 2023, malicious actors have sent text messages and AI-generated voice messages” claiming to come from senior officials to build rapport with targets, then moved conversations to encrypted messaging apps. According to the FBI, the actors have asked victims for authentication codes, personal information and copies of documents such as passports, wire transfers to overseas financial institutions, and introductions to associates. The FBI says AI-generated voices “can sound nearly identical” to a known contact and advises verifying identities through an independently found phone number.

On September 10, 2026, Microsoft described a campaign it detected between August 3 and 5 that sent “more than a million” emails, 87.7% of them to users in the United States. Microsoft says the actor impersonated executives of the targeted companies, such as a CEO, CFO, or president, to ask accounts payable teams to process an ACH payment of nearly $50,000. The emails included a fabricated invoice and a fake forwarded email thread with an impersonated software vendor, and used lookalike domains. Microsoft says it saw “several indicators consistent with AI-assisted template development,” but adds that “while these indicators suggest generative AI involvement, they do not independently establish the extent to which AI generated campaign content.” Microsoft says it found no evidence that the legitimate organizations referenced in the lures were compromised.

Why it matters to leaders (our analysis)

Both cases rely on the target trusting a familiar name, title, or voice. The FBI says AI-generated voices can sound nearly identical to a known contact, and Microsoft saw signs of AI-built email templates, so recognition alone is a weak control. The fixes are procedural: verify through a separate, known channel, require two people for unusual payments, and give no exceptions to requests that appear to come from the top.