OpenAI released GPT-6 Cyber — a model specifically designed to find and exploit security vulnerabilities, to a small group of audited customers, and rated its cybersecurity risk rating as “critical” based on its security framework. OpenAI said that given the increasing number of attempts by malicious cyber attackers to abuse AI, the model aims to help security researchers find bugs, test how they can be exploited, and fix these flaws before attackers get their hands on them.

OpenAI released GPT-6 Cyber — a model specifically designed to find and exploit security vulnerabilities, to a small group of audited customers, and rated its cybersecurity risk rating as “critical” based on its security framework. OpenAI said that given the increasing number of attempts by malicious cyber attackers to abuse AI, the model aims to help security researchers find bugs, test how they can be exploited, and fix these flaws before attackers get their hands on them.