#AI Hacker breaches 50% #DeFi contract! #Anthropic report warning, security battle enters AI confrontation era
On December 4, 2025, #Anthropic released a red team report revealing fatal risks: top AI models (such as Claude Opus 4.5) can autonomously exploit DeFi smart contract vulnerabilities, successfully breaching 50% of previously unseen new contracts during testing, potentially siphoning off 4.6 million dollars within the simulated attack window. Even more terrifying is that the cost of scanning a single contract is only 1.22 dollars, yet it can leverage thousands of dollars in returns. After a failed attack, it will automatically adjust parameters and retry. This means traditional manual auditing is gradually becoming ineffective; the previous liquidity crisis of #Yearn.Finance has exposed the vulnerabilities in DeFi, and now AI attacks have significantly raised the security threshold for the industry. Currently, leading protocols are urgently deploying AI defense systems to address the crisis by monitoring contract vulnerabilities in real-time and automatically repairing code. $YFI
Do you think AI defenses can withstand AI attacks? Share in the comments which DeFi protocol security issues you are most concerned about～
