Note that the ChatGPT bot’s capabilities have previously been used for the cryptocurrency industry. In the middle of the month, Cointelegraph journalists forced the artificial intelligence to examine recent articles, then asked it to share a list of cryptocurrencies to invest $100.

In the end, half of the amount the bot offered to allocate to Bitcoin, while $25 and $15 went to Etherium and Cosmos, respectively. The remaining amount was split between MANA and LINK tokens.

However, this experiment has had its fair share of drawbacks. First of all, the artificial intelligence cannot predict the future, so its advice is just a conclusion from the studied materials. In addition, the articles offered to the bot do not cover everything that is happening in the digital asset market, so even this artificial intelligence solution will be limited.

Be that as it may, it was fun to evaluate ChatGPT’s capabilities anyway. That’s why Coinbase employees used it as well.

How is ChatGPT bot used?

A note about the Coinbase experiment appeared on the exchange’s official blog. Here’s how analysts described the conditions posed to the artificial intelligence.

If we simply asked ChatGPT to produce a risk assessment based on the security of a particular token’s smart contract, the tool wouldn’t have enough contextual information to produce a response that we could compare to manual verification. This is because ChatGPT and a live auditor would not evaluate the security of smart contracts in the same way.


Indeed, when interacting with artificial intelligence, it is crucial to specify the terms of the task. As practice shows, it is necessary to gradually add new details to solve a certain task in order to get the desired result. Even when it comes to generating a picture, it is worth looking at the intermediate outcome of the artificial intelligence and then sharing the new commands and conditions for the task.

The very evaluation of a certain token's smart contract is a necessary step before adding a crypto-asset to an exchange. The procedure also allows one to examine the permissions that are required to interact with a smart token. This will allow you to make sure that it has not, for example, been created by the developers solely for the purpose of stealing users' money.


To make the results more like an audit, ChatGPT initially gave enough information about the algorithm used by Coinbase analysts to evaluate smart contracts. The artificial intelligence then examined twenty smart token contracts, and in twelve of those cases, ChatGPT’s results were almost exactly the same as those of the auditors.

At the same time, among the eight cases where ChatGPT got it wrong, five of the results were critical blunders by the artificial intelligence. In them, it marked high-risk assets as safe, which is unacceptable for audit results.

Variability is also an issue with artificial intelligence, as it can produce different results for the same query after a period of time. However, the Coinbase team is optimistic about the results of their experiment. ChatGPT could be used as an additional tool for auditors to double-check their reports, experts say.

An analyst nicknamed Officer’s Notes in an interview with Decrypt supported Coinbase experts’ optimism. Here’s his quote.

At this stage, artificial intelligence cannot replace humans, but it is an indispensable assistant – including for tired or inattentive auditors. I think quality assurance and phasing will definitely not be able to do without artificial intelligence tools in the future.

A speaker from BlockSec, a blockchain cybersecurity company, also commented on the Coinbase initiative. Here’s the rejoinder.

We are not surprised because such smart contracts can also be automatically verified by other traditional programming tools. However, artificial intelligence cannot run on complex business logic, which is a major attack vector and a major loophole that smart contracts should focus on auditing.

???? YOU CAN FIND MORE INTERESTING STUFF ON US AT YANDEX.ZEN!

Can artificial intelligence fully replace human auditors in the foreseeable future? Connie Lam, Head of Solutions Architecture at Certik, shares his opinion on this.

We may get to that point at some point, but not in the near future. A complementary approach is more likely. There are things that humans do better than machines, and vice versa. Tools help us create new things, but they don’t replace us. The invention of the calculator did not make accountants obsolete, it made them better at their jobs.

Indeed, it is conceivable that, in the future, artificial intelligence will become an assistant to auditors who study smart contracts and other blockchain platforms. It could, for example, be tasked with routine tasks such as looking for dangerous approvals and the like.


Artificial intelligence seems to be actively starting to gain new ground, and the cryptocurrency industry recognises this. As can be understood from the Coinbase experiment, ChatGPT is already showing good results in complex operations. And as the platform improves, its capabilities are sure to expand, which will also benefit the digital asset industry.