OpenAI近日推出了一款专注于网络安全领域的新模型“ChatGPT Cyber”,目前该模型已向参与Daybreak Red计划的高级用户开放 1。为了强化关键基础设施的防御能力,OpenAI同时宣布将投入10亿美元专项资金,用于支持该领域的网络安全专业人员 1。
在新模型的应用成效方面,OpenAI表示ChatGPT Cyber已被用于扫描Linux内核和Firefox浏览器,分别识别出948个潜在问题和81个新漏洞 1。此前,Charge Robotics首席技术官Max Justicz利用OpenAI的GPT-5.6 Sol模型,在短短2小时内发现了Ubuntu系统中一个严重的服务器控制漏洞 1。针对此次Linux漏洞猎捕行动,总成本为120万美元,其中Justicz个人花费约5000美元用于订阅费和Token费 1。OpenAI研究员Eric Wallace进一步透露,公司正分配大量前沿算力,免费用于开源代码库的漏洞发现与修复工作 1。
OpenAI has introduced ChatGPT Cyber, a new model dedicated to cybersecurity, which is currently available to advanced users participating in the Daybreak Red program 1. The company also announced a $1 billion initiative designed to support cybersecurity professionals working in critical infrastructure sectors 1. This launch follows demonstrations of AI's efficacy in identifying severe security flaws, including an incident where Charge Robotics CTO Max Justicz used the GPT-5.6 Sol model to discover a serious server control vulnerability in Ubuntu within just two hours 1.
According to OpenAI, the new model has already been utilized for large-scale code scanning efforts with significant results. During a Linux vulnerability hunt that cost $1.2 million, the model identified 948 potential issues in the Linux kernel and 81 new vulnerabilities in the Firefox browser 1. Justicz personally spent approximately $5,000 on subscription and token fees during his investigation 1. OpenAI researcher Eric Wallace stated that the company is allocating substantial frontier compute resources free of charge to assist in finding and fixing vulnerabilities in open-source codebases 1.
评论
还没有评论,欢迎留下第一条。