gpt.buzz
Sign in

news

Introducing EVMbench

February 18, 2026

OpenAI and Paradigm introduced EVMbench, a benchmark for evaluating AI agents’ ability to detect, patch, and exploit high-severity smart contract vulnerabilities. It matters because smart contract security is a high-stakes domain where agent performance can be measured on realistic attack, defense, and remediation tasks.

OpenAI and Paradigm introduce EVMbench, a benchmark evaluating AI agents’ ability to detect, patch, and exploit high-severity smart contract vulnerabilities.

Source: openai.com

← All news