Generative artificial intelligence (AI) technology such as OpenAI’s ChatGPT-4 is not yet able to audit smart contracts as effectively as human auditors, according to blockchain security firm OpenZeppelin. AI researchers at the company challenged the bot to hack Ethernaut security, and while it passed the majority of the levels, it struggled with newer ones that had been introduced since its September 2021 training data cutoff date. Allowing a web connectivity plugin would have assisted the bot in solving these issues. The researchers concluded that even though ChatGPT-4 cannot replace human auditors, it can be used to detect security vulnerabilities and boost the efficiency of smart contract auditors.
OpenZeppelin is a blockchain security company that provides security audits of smart contracts, crypto holdings, and customer-facing applications.
Mariko Wakabayashi and Felix Wegener are researchers at OpenZeppelin who pitted ChatGPT-4 against the Ethernaut security challenge to test its ability to audit smart contracts.
