AI-Powered Crypto Phishing: Economics Shift as Attacks Scale

The unit economics of crypto phishing have changed. Attackers can now generate convincing lures at near-zero marginal cost and personalize them at scale. The data shows rising average payments and explosive growth in impersonation scams, alongside AI's integration with phishing-as-a-service kits.

Chainalysis estimates at least $14 billion flowed to on-chain scam addresses in 2025, with total scam and fraud losses potentially reaching $17 billion. The average scam payment jumped 253% year over year from $782 in 2024 to $2,764 in 2025. Impersonation scams grew about 1,400% year over year. AI-enabled scams were roughly 4.5 times more profitable than traditional scams, the firm said.

The FBI's IC3 recorded 1,008,597 total complaints in 2025 and flagged 181,565 cryptocurrency-related complaints with reported losses of $11.366 billion. The bureau also tracked AI-related complaints for the first time, logging 22,364 complaints and $893.346 million in reported losses, its annual report said.

Barracuda's analysis of more than 3.1 billion emails found one in three were malicious or unwanted. Phishing accounted for 48% of malicious email, and 90% of high-volume phishing campaigns used phishing-as-a-service. Adversaries are pairing AI-driven social engineering with PhaaS to scale targeted credential harvesting, Barracuda said.

The combination of content-generation AI with turnkey delivery infrastructure is the material change. Language and image models reduce the time and skill needed to write targeted copy, build fake support chats, and fabricate identity documents. PhaaS kits provide distribution, hosting, templates, and credential-stealing logic that historically required bespoke effort.