Anthropic tested ten AI models on 405 historical smart contract exploits, reproducing 207 of them, indicating that AI can effectively exploit significant vulnerabilities.
The AI models collectively simulated potential exploit revenues of $550 million, demonstrating their ability to match skilled human attackers in exploiting smart contract vulnerabilities.
Two new zero-day vulnerabilities were discovered on Binance Smart Chain contracts, underscoring the risk of AI-enabled attacks.
Experts warn about the misuse of AI in malicious attacks, emphasizing the need for robust security measures to counteract these risks.
Leave a Reply