AI Exploits Pose Growing Risk to Smart Contracts

[johnblockbuster]

Adding to crypto security concerns, recent research shows AI models can autonomously find profitable smart contract exploits. Testing by Anthropic and MATS revealed AI tools like Claude 4.5 and GPT-5 could exploit contracts to generate $4.6M in value, even discovering zero-day flaws in recently deployed contracts. Combined with attacks like the whale’s multisig wallet hack, these findings highlight the rising need for advanced security practices in crypto, from wallet configuration to contract auditing, as both human and AI attackers grow more sophisticated.