Odaily Planet Daily News An executive from the security agency OpenZeppelin said that a recent company experiment proved the continued need for human auditors. OpenZeppelin conducted a study testing GPT-4’s ability to identify various smart contract vulnerabilities across 28 Ethernaut challenges. Of the 23 Ethernaut challenges introduced before GPT-4’s training data deadline (September 2021), the AI ​​tool successfully solved 19 and then failed in 5 challenges introduced after September 2021. 4. In this regard, OpenAI pointed out on its website that this AI tool "generally lacks" understanding of events that occurred after September 2021 and "cannot learn from experience."