Coinbase, a digital asset exchange, has tested the artificial intelligence ( AI) chatbot created by Openai for automated token reviews. According to the U.S.-based trading platform, Chatgpt wasn’t accurate enough to be integrated immediately into its asset review process but it had enough potential to merit further investigation.

This experiment is part Coinbase’s effort to use efficient and effective methods for reviewing token contracts before listing the assets. The exchange explained that the Blockchain Security team uses in-house automation tools to assist security engineers in reviewing ERC20/721 smart contract contracts.

ChatGPT by OpenAI has been a hot topic due to its ability to detect security flaws. We wanted to see if this could be used as a frontline tool and not just for code review.

Coinbase explained that Chatgpt had shown promise in improving productivity across a wide variety of engineering and development tasks. The AI tool can also be used to optimize code or identify vulnerabilities.

A leading American cryptocurrency exchange conducted an experiment to determine if Chatgpt’s token security review was as accurate as a standard review done by a blockchain security engineer using their internal tools. The chatbot needed to be trained to recognize risks according to the platform’s security review framework in order to produce comparable risk scores.

Researchers compared 20 smart contracts risk scores between Chatgpt, and a manual security inspection. The AI tool produced identical results to the manual review 12 more times. Five of the eight missed cases were instances in which Chatgpt mistakenly classified a high-risk asset with a low-risk one. The exchange stated in a blog that underestimating a risk score can be more damaging than overestimating it.

Coinbase claims that Chatgpt’s efficiency has been amazing, despite the tool’s inconsistencies and ‘worst-case failure’. The company believes that the accuracy of the tool could be improved with prompt engineering.

Coinbase stated that the bot can’t be relied on to do a security review. It also stated that the bot can’t be relied upon to perform a security review. However, its team could improve the accuracy. It means engineers could potentially use it to perform additional controls checks to detect potential risks that might have been missed.

Openai’s Chatgpt platform is in the spotlight due to growing popularity of artificial Intelligence applications. Binance, the largest cryptocurrency exchange in the world, had announced that it was launching an AI-centric token ( NFT ).