Could AI Agents Exploit Ethereum, XRP, Solana? Anthropic Says It's Possible
-
最新报道：Anthropic's latest tests reveal AI agents autonomously exploited smart contracts, draining millions in simulated funds. In one test, models like Claude Opus 4.5 successfully attacked 17 of 34 new contracts, generating $4.5 million. In a broader experiment on 405 previously exploited contracts across Ethereum, BNB Chain, and Base, AI executed 207 profitable attacks for a simulated $550 million. The agents replicated real hacker behavior, identifying bugs and scripting full exploits. Researchers also found GPT-5 and Sonnet 4.5 could discover new, zero-day vulnerabilities. Anthropic warns that as AI model costs decline, automated, large-scale blockchain exploitation becomes a significant threat, though the same technology could be adapted for defense.
下一篇
Bybit, in partnership with Mantle and Aave, launched a global institutional-grade on-chain liquidity network.
- Bybit has announced a strategic partnership with Mantle and Aave. The three parties will j...
24小时热点
热点专题
407042