On April 14, 2026, the Coinbase exchange announced the development of Frosty, an artificial intelligence tool designed to audit smart contracts, which is already part of its internal security review process.
According to the company, the system was evaluated in parallel with six other AI tools. We analyzed 33 real audits with 434 verified vulnerabilities. In these tests, Frosty obtained better results in metrics used to measure the effectiveness of fault detection such as accuracy, coverage, and F1 score.
The tool works through an autonomous architecture based on: Multiple agents and consecutive phases. This process includes tasks such as code analysis, searching for vulnerabilities, adversarial inference to simulate possible attacks, debugging the results, and generating preliminary reports. Each run takes 1-2 hours and generates a report that is then reviewed by a human team.
According to the company, To implement this type of system, you need to: Toward expanding the use of artificial intelligence Both by developers and potential attackers. In this context, automation tools aim to speed up the detection of errors in the early stages of development.
However, Coinbase says that Frosty Not a replacement for traditional audits conducted by experts. Its use is intended as a complement to the review process, as this tool may overlook complex or situational vulnerabilities.
The development of this type of solution will take place in parallel with other efforts in this field. for example, OpenAI recently introduced EVMbenchAs reported by CriptoNoticias, it is a test environment that measures the performance of artificial intelligence agents in detecting, fixing, and exploiting smart contract failures. Although these tools represent progress, results vary depending on the task.
(Tag translation) Coinbase
