SYNTHETIC DECEPTION: DEEPFAKES IN ELECTIONS, VOICE SCAMS, AND IMAGE ABUSE
The dollar losses get headlines, but the deepest harms of synthetic media land on people: voters fed fake robocalls, families drained by cloned voices, and victims — overwhelmingly women — targeted by nonconsensual imagery. This is the human-facing edge of AI’s dark side, and after years of lagging, the law is finally moving.
Thank you for reading this post, don't forget to subscribe!
Key takeaways
- →Elections: an AI-voiced “Biden” robocall told New Hampshire voters to stay home; the operative drew a $6M FCC fine. (FCC)
- →Scams: voice clones now need as little as three seconds of audio, and one in four adults reports encountering an AI voice scam. (McAfee)
- →Abuse: roughly 96% of all deepfakes are nonconsensual intimate imagery, 99% of it targeting women; AI child-abuse videos exploded in IWF data. (Sensity / IWF)
Elections and disinformation
Two days before the January 2024 New Hampshire primary, thousands of voters got a robocall in President Biden’s AI-cloned voice urging them not to vote. The FCC issued a $6 million forfeiture against the consultant behind it and pursued the transmitting carrier, Lingo Telecom, which settled for $1 million. With dozens of countries holding elections in 2024, identity-verification firms reported sharp spikes in political deepfakes — making provenance and rapid debunking part of election security.
Voice-clone and impersonation scams
The same technology hits households. A clone convincing enough to fool family can be built from a few seconds of audio scraped from social media, powering “grandparent” and CEO-impersonation scams. A McAfee study found one in four adults had experienced an AI voice scam, and FTC imposter-scam losses reached roughly $2.95 billion in 2024. The defense is low-tech: a family code word, and a hang-up-and-call-back rule for any urgent money request.
“The deepest harms of synthetic media land on people — and the defense is often as low-tech as a family code word.”
Image-based abuse: the gravest harm
The most disturbing trend is the use of AI to generate abuse imagery. The Internet Watch Foundation reported AI-generated child sexual abuse videos rising from 13 in 2024 to 3,440 in 2025, most rated in the most severe category — material now producible at scale with little technical skill. More broadly, the overwhelming majority of deepfakes online are nonconsensual sexual imagery, almost entirely targeting women. This is the harm driving the fastest legislative response.
The legal response
Lawmakers are catching up. The US TAKE IT DOWN Act, signed in May 2025, made publishing nonconsensual intimate imagery — explicitly including AI “digital forgeries” — a federal crime, and around 45 states now have deepfake laws covering elections, abuse imagery, or voice cloning. In the EU, the AI Act will require AI-generated content to be labeled and machine-detectable from August 2026, with penalties reaching €35M or 7% of global turnover for the most serious violations.
Frequently asked
Are deepfakes actually affecting elections?
Yes. The 2024 AI-voiced Biden robocall that urged voters to skip the New Hampshire primary led to a $6M FCC fine, and verification firms recorded surges in political deepfakes across the many countries holding elections that year.
Is creating or sharing deepfakes illegal?
Increasingly, yes. The US TAKE IT DOWN Act (2025) criminalized nonconsensual intimate deepfakes federally, about 45 states have deepfake laws, and the EU AI Act will require AI content to be labeled from 2026 with heavy penalties for violations.
The AI Index (2026). Synthetic Deception: Deepfakes in Elections, Voice Scams, and Image Abuse. Retrieved Jun 20, 2026, from report-ai.org/reports/dark-side-of-ai/ai-deepfake-disinformation-voice-scams/