SlowMist Annual Security Report: Total security incidents resulted in approximately $2.935 billion in losses, with generative AI frequently used as an attack tool

On December 30, according to incomplete statistics from SlowMist Blockchain, a total of 200 security incidents occurred throughout the year, resulting in losses of approximately $2.935 billion. Compared to 2024 (410 incidents, approximately $2.013 billion in losses), although the number of incidents has significantly decreased, the amount of losses has increased by about 46% year-over-year.

From an ecosystem distribution perspective, Ethereum remains the most frequently attacked and most severely affected ecosystem, with losses of about $254 million for the year, significantly leading in share; BSC follows closely, with related losses of approximately $21.93 million; Solana ranks third, with annual losses of about $17.45 million.

Dividing by project track, DeFi projects are the most commonly targeted: in 2025, there were 126 security incidents, accounting for about 63% of the total for the year, causing losses of approximately $649 million, a decrease of about 37% compared to 2024 (339 incidents, $1.029 billion in losses).

There were only 12 exchange platform incidents, but they caused losses as high as $1.809 billion, with a single CEX incident losing about $1.46 billion, making it the most severe event of the year.

Regarding the causes of attacks, contract vulnerabilities are the main trigger, with 61 incidents; account hacking (X account being compromised) follows closely behind, with 48 incidents. The report points out that as generative AI has accelerated its adoption over the past two years, attackers have also begun to incorporate it into scams and attack chains. Compared to traditional tools, AI significantly reduces the cost of scams in text, voice synthesis, image, and video generation. Attacks no longer rely on crude language or obviously abnormal behavior but instead use highly realistic content, coherent interactions, and precise targeting to make it more difficult for victims to perceive the risks psychologically.