digital democracy|dystopian new tech
$3.7B lost to deepfakes, social media is the primary origin
Deepfakes are no longer just a novelty — they are a constant threat hidden in everything from social media posts to phone calls. In the key insights below, we break down where this threat is growing fastest and where it most often begins, also, how much money it has cost so far, and why it can no longer be ignored.
Key insights
- Deepfake fraud is accelerating sharply, with recorded global losses now reaching $3.7 billion. Losses totaled just $83 million between 2020 and 2023, then rose to $335 million in 2024, surged to $2.5 billion in 2025, and already reached $764 million in the first half of 2026. Together, 2025 and 2026 account for 89% of all recorded global deepfake fraud losses, showing how quickly the threat has escalated.
- Social media is the single largest origin point for deepfake-related losses, accounting for $1.73 billion, or 47% of the global total. Losses originating on these platforms are driven largely by criminals using deepfakes of celebrities and public figures to promote fraudulent investment schemes.
- Impersonation fraud is the second-largest origin category, accounting for $911 million, or 25% of losses. This category includes cases where criminals use deepfake-enabled tactics to impersonate real people, bypass identity checks, or gain unauthorized access to financial services. Common examples include synthetic facial verification videos, AI-generated face swaps, voice cloning, forged or altered IDs, fraudulent account openings, loan application fraud, and wallet or banking access scams.
- Crypto ATM fraud generated $333 million in losses in 2025, making it one of the largest emerging deepfake fraud types originating on social media. In these cases, scammers use AI-powered deepfakes to pose as trusted individuals and pressure victims, especially older adults, into sending money through fast and anonymous cryptocurrency ATMs.
- Fake job candidate schemes added another $100 million in losses, as criminals used AI-generated resumes, face-swapping, and deepfake video interviews to infiltrate hiring processes, particularly for tech roles in the US.
- Sadly, deepfake fraud is also moving into everyday communication channels. Phone calls generated $71 million in losses, making them the largest channel in this group, followed by video platforms and conferencing tools at $62 million, and messaging apps at $41 million. Email contributed a smaller but still notable $12 million, showing that deepfake scams are no longer limited to major social media platforms, but are increasingly reaching victims through routine voice, video, messaging, and email interactions.
Methodology and sources
This study used data from the AI Incident Database, Resemble.AI, and the OECD to create a combined dataset covering deepfake incidents from January 2020 to June 2026. Incidents were included if they involved the generation of synthetic videos, images, or audio; were verified by media reports; and had clearly documented financial losses.
Each deepfake incident was classified by target country and origin. These figures represent a conservative estimate based on publicly reported data, and the source databases included records of deepfake incidents across multiple languages.
If multiple countries were mentioned in an incident with a specific financial loss, that loss was divided equally among the countries involved.
To avoid double counting, each incident was assigned to one primary origin category based on the main channel, tactic, or entry point described in the source material. The categories were defined as follows:
- Social media: cases originating on social media platforms;
- Impersonation fraud: cases involving impersonation, synthetic facial verification videos, AI-generated face swaps, voice cloning, forged or altered IDs, fraudulent account openings, loan application fraud, or unauthorized access to wallets, bank accounts, or financial services;
- Website: cases originating from fraudulent websites, including fake investment platforms, scam landing pages, or other web-based fraud schemes;
- Fake job candidates: cases involving AI-generated resumes, face-swapping, deepfake video interviews, or other synthetic identity tactics used to infiltrate hiring processes;
- Phone call: cases where deepfake-enabled voice cloning or impersonation was used during phone-based scams;
- Video platform: cases originating on or using video platforms, including video conferencing tools or video-sharing platforms;
- Messaging app: cases involving deepfake-enabled scams carried out through messaging applications;
- Email: cases involving deepfake-enabled fraud delivered or initiated through email;
- Other: cases that do not fit into the above categories, including novel schemes or unclassified methods.
