Anthropic: AI agent discovers $4.6 million vulnerability in real-world contracts
-
最新报道：According to a report by Anthropic, its researchers tested Claude Opus 4.5, Claude Sonnet 4.5, and GPT-5 models on their self-built SCONE-bench benchmark (containing 405 real-world attacked contracts from 2020 to 2025). In contracts attacked after the knowledge update date (March 2025), they discovered exploitable vulnerabilities worth approximately $4.6 million. Furthermore, in simulation tests of 2,849 recently deployed contracts without known vulnerabilities, Sonnet 4.5 and GPT-5 each discovered two new zero-day vulnerabilities, potentially causing a total loss of $3,694, with GPT-5's API costing $3,476.
下一篇
Today in History | December 2
- According to CoinWorld, on this day in 2023, Adam Weitsman purchased Gold Mega Koda #9969 ...
24小时热点
热点专题
407106