Blog Post

Releasing Our Digital Asset Risk Annual Report - The Rise of AI-Drive Crypto Fraud

January 15, 2026

AI Didn’t Just Accelerate Crypto Fraud in 2025 — It Industrialized It

AnChain.AI - Jan 2025

‍

In 2025, crypto fraud crossed a critical threshold. It stopped being opportunistic and became systematic, automated, and AI-orchestrated.

At AnChain.ai, we analyzed the top AI-driven crypto fraud incidents of 2025—cases where artificial intelligence was not incidental, but central to how the crime was planned, executed, and scaled.

The result is our 2025 AI Risk Report, and the findings are sobering.

Article content

$2+ Billion Lost — From The Top 10 AI-Driven Incidents

Across just 10 major AI-enabled crypto fraud cases, more than $2 billion was stolen.

Article content

These include:

  • Bybit cold-wallet hack — ~$1.5B
  • Cetus DEX overflow exploit — ~$223M
  • 783-BTC deepfake support scam — ~$91M
  • Phemex hot-wallet breach — ~$70M
  • BtcTurk multi-chain hack — ~$48M
  • SwissBorg / Kiln supply-chain attack — ~$42M
  • GMX V1 re-entrancy exploit — ~$42M
  • UXLINK multisig key theft — ~$41M
  • ALEX Protocol vault exploit — ~$16M
  • Resupply donation manipulation — ~$9.5M

These are only a subset of 2025 incidents, but they clearly expose how AI is reshaping financial crime.

How AI Was Used — At Scale

What matters more than the individual hacks is the pattern behind them:

Article content
  • Large Language Models (LLMs) were used in 7 of the 10 cases, enabling phishing, fake support chats, and approval manipulation—linked to ~$1.8B in losses.
  • Automated bots appeared in 6 cases, powering flash-loan orchestration, oracle manipulation, and rapid fund movement—nearly $1.9B in impact.
  • AI-assisted code analysis and exploit generation showed up in 5 cases, accelerating smart-contract exploitation—~$1.8B affected.
  • Voice and video deepfakes, while used in fewer cases, still contributed to $1.5B+ in losses by exploiting human trust rather than technical vulnerabilities.

These numbers overlap because modern attacks are AI-stacked—language, identity, code, and automation operating together as a single fraud system.

Fraud Is Now Machine-Scale

The most damaging incidents of 2025 shared a common architecture:

LLMs generate the narrative → Voice/video AI establishes authority → Automated tools exploit logic → Bots execute and launder funds at machine speed.

This is no longer “fraud with AI tools.” It is AI as the operating system of financial crime.

Why This Matters for 2026

AI-driven fraud doesn’t just exploit vulnerabilities—it exploits latency:

  • Latency in human review
  • Latency in rule-based AML systems
  • Latency between detection and response

As long as attackers operate autonomously and defenders operate manually, the gap will keep widening.

The Case for Super-Intelligent Defense

Our conclusion is straightforward:

When fraud operates at machine speed, defense must do the same.

Protecting crypto and fintech ecosystems now requires super-intelligent systems—platforms that can reason, correlate, and act across chains, identities, and behaviors in real time, while remaining explainable for regulators and law enforcement.

Read the Full Report

The AnChain.ai 2025 AI Risk Report includes:

  • Deep dives into each of the top 10 incidents
  • A technical breakdown of AI capabilities used by attackers
  • Quantitative loss analysis by AI category
  • Practical recommendations for institutions and policymakers

👉 Download the full report: https://www.anchain.ai/2025-risk-report

‍

Sign Up For Our Newsletter

Thank you for signing up!
Something went wrong! Please try again.

© 2025 AnChain.AI. All Rights Reserved | Privacy Policy