AI is reshaping the on-chain offense and defense landscape in crypto: monthly frequency of on-chain attacks has risen to 2.8 times the previous rate in the past two years, significantly increasing protocol security pressure.

BlockBeats news, on June 5, according to BlockBeats statistics, after the widespread adoption of AI, the monthly frequency of on-chain attacks is approximately 2.8 times greater than before. Using the period from 2020 to 2021 as the baseline prior to large-scale AI utilization, the monthly average number of on-chain attack incidents was about 3.7; while in the most recent two years from June 5, 2024 to June 5, 2026, the monthly average was about 10.4 incidents, totaling 250 incidents.

Looking at annual data, 17 on-chain attacks occurred in 2020, 71 in 2021, and 67 in 2022. After the release of ChatGPT, the number increased to 97 in 2023, 94 in 2024, 97 in 2025, and already 96 in 2026 (annualized about 225 incidents). The data shows that the number of attack incidents has remained consistently above the early baseline since 2023.

High-frequency attack windows were mainly concentrated from the end of 2025 to 2026. Specifically, in the 56 days after the release of GPT-5.2 on December 11, 2025, there were 24 attacks; in the 42 days after the release of GPT-5.4 on March 5, 2026, there were 21 attacks; and in the 35 days after the release of GPT-5.5 on April 23, 2026, there were 38 attacks, which is the highest count in the chart.

The frequency of on-chain attacks increased significantly following AI’s widespread adoption, and has remained at a high level during multiple recent model release windows, indicating that crypto security defense and offense are entering a more frequent and automated stage. With advancements in AI-assisted auditing, vulnerability detection, and attack script generation, the importance of protocol security, smart contract auditing, and real-time monitoring continues to rise.