Older Americans lost at least $1.9 million to grandparent scams in the first nine months of 2023 alone, according to FBI complaint data, and federal regulators now warn that the same scheme has gained a dangerous new weapon: AI-generated voice cloning that can mimic a family member’s speech from a handful of audio clips posted online. The shift turns a decades-old confidence trick into something far harder to detect, because the voice on the other end of the line sounds exactly like a grandchild, niece, or nephew in distress.
How cloned voices supercharge a familiar con
The basic script has barely changed in years. A caller poses as a relative in legal trouble or physical danger, pressures the target to send cash immediately, and insists on secrecy. The Department of Justice has indicted multiple defendants in organized grandparent-scam networks on racketeering conspiracy charges, showing how the scheme operated at industrial scale even before artificial intelligence entered the picture.
What has changed is the opening seconds of the call. The Federal Trade Commission explains in a consumer alert that scammers now clone a loved one’s voice from real audio found on social media or other public sources. A few seconds of a birthday video, a podcast clip, or a voicemail greeting can be enough. The cloned output is then fed into a live call so the target hears what sounds like their own grandchild pleading for help. The FTC puts it bluntly: “Voice cloning, that’s how.”
The FBI has separately warned that criminals are using generative AI to lower the technical barrier for financial fraud across the board. An FBI IC3 public service announcement details how generative AI tools reduce the skill needed to produce convincing audio, text, and imagery at scale. For grandparent scams specifically, that means a single operator can now run dozens of personalized calls per day rather than relying on a generic acting performance that many seniors would recognize as fake.
FBI and FTC complaint data behind the $1.9 million toll
Between January and September 2023, the FBI’s Internet Crime Complaint Center logged more than 195 complaints tied to grandparent scams, with combined reported losses of at least $1.9 million. Those figures almost certainly undercount the real damage. Many victims never file a report, either because they feel embarrassed or because they do not realize a crime occurred until weeks later. The IC3 alert, designated I-111723-PSA, describes the operational mechanics in detail: callers create urgency, demand wire transfers or cash sent by courier, and coach victims to lie to bank employees about the purpose of the withdrawal.
A separate FBI cyber alert states that malicious actors increasingly use AI-generated audio for voice-phishing attacks, including impersonation of “personal relations.” While that alert focused on campaigns targeting senior government officials, the technique is identical to what hits private citizens in grandparent scams. The difference is that a retired teacher in Ohio has far fewer resources to verify a suspicious call than a federal employee with an IT security team.
The FTC treated voice cloning as a serious enough consumer threat to launch a formal Voice Cloning Challenge, and in April 2024 the agency announced the challenge winners. The competition asked technologists to propose ways to detect or block cloned audio before it reaches a victim. The FTC’s own technical staff published an accompanying analysis noting that voice samples are easy to obtain and that current defenses remain limited. That gap between offensive capability and defensive tooling is the core reason the threat is growing faster than the response.
Why existing IC3 data cannot yet map the AI-specific risk
One significant blind spot in the public record is the absence of complaint data broken down by whether a call used AI-generated audio or a live human impersonator. The IC3’s 195-plus complaints and $1.9 million loss figure cover grandparent scams broadly. Neither the FBI nor the FTC has released a subset count isolating AI-cloned-voice incidents from traditional ones. That means analysts cannot yet measure how much of the recent growth in losses is driven by cloning versus the same old playbook.
A testable pattern may emerge as more data becomes available. Seniors whose social-media accounts contain short, publicly accessible voice clips, such as Facebook video greetings or TikTok narrations, present a richer source library for scammers building cloned-voice profiles. If the IC3 eventually publishes complaint data at the zip-code level, researchers could cross-reference those hotspots with platform metadata on audio-post density to see whether regions with more vocal social-media users report higher rates of AI-enhanced fraud. That analysis does not exist yet, but the building blocks are already in federal databases.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
