TRM Labs says AI use in crypto crime rose 40% in a year
The firm's new index puts AI adoption across crypto crime at 54 out of 100, with scams the most advanced category and deepfake losses already above 2025's full-year total.
By The Third AnglePublished 3 min read
Illustrative software-security imagery; it does not depict TRM Labs, a criminal operation or a specific incident. Photo: Unsplash · Unsplash License
Artificial intelligence adoption across crypto crime rose 40% over the past year, according to a new TRM Labs report. The firm's 2026 AI-in-Crime Adoption Index scores overall use at 54 out of 100, up from about 28 in 2024, and places scams at the most advanced stage of adoption.
The Block reported the findings on Aug. 21, citing TRM's new index. The figures describe TRM's methodology and dataset, not an independent audit of every crypto crime report. The report says its index combines prevalence, the breadth of AI use across a crime lifecycle and the sophistication of the tools involved.
Scams lead the adoption curve
TRM says the share of scam reports involving deepfakes, AI chatbots or AI-branded lures has grown roughly 13 times since 2022. Reported losses from deepfake scams in 2026 to date have already exceeded the full-year 2025 total by 263%, according to the report.
The important distinction is between criminals using AI and victims using consumer tools after a scam. TRM says about half of the increase in its measure comes from scams in which AI was part of the attack, while the remainder reflects victims using tools such as ChatGPT to investigate activity they had already encountered.
Hacks and ransomware are catching up
The report says crypto hacks reached a record 201 incidents in the first half of 2026, more than double the comparable 2025 figure. About 4% of incidents accounted for 75% of losses, and TRM says most of those large incidents involved infrastructure compromises such as private-key or credential theft. North Korea-linked activity accounted for about $600 million, or 61% of first-half losses, according to the report.
TRM also describes ransomware as an emerging AI-use category, with no-code kits selling for $400 to $1,200 and a July operation called JadePuffer presented as the first documented agentic ransomware attack. Those claims point to a changing threat model, but they do not show that every attack is autonomous or that AI replaces human operators across the wider market.
