OpenAI's autonomous pipeline finds and patches 948 Linux kernel privilege escalation bugs
Article timeline
The articles in this story. Times are ET.
Greg BrockmanOfficialPickOpenAI allocates frontier compute to finding and patching open-source vulnerabilitiesAIOpenAI says it is allocating a significant amount of its frontier compute to discovering and patching vulnerabilities in open-source codebases, fully pro bono. In one case, its upcoming frontier cyber model found about 1000 local privilege escalation vulnerabilities in the Linux kernel in one week using about $1M of tokens. The post says the model also discovered at least 10 novel classes of Linux exploitation techniques, compared with roughly 20 such vulnerabilities a typical year before AI.
- Greg BrockmanOfficialPickOpenAI reports patches for almost 1,000 Linux privilege escalation flaws
AIOpenAI says its cyber vulnerability pipeline found, validated, and prepared patches for almost 1,000 Linux local privilege escalations, at about $1,000 per vulnerability. A quoted OpenAI post says one run with an upcoming cyber model validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel, at a total API cost of $1.2 million. The accompanying funnel shows 2,308 candidate findings, 1,724 validated findings, and 948 working local privilege escalation exploits.
Yuchen JinXOpenAI reports 948 Linux kernel privilege escalation exploits from an autonomous pipelineAIOpenAI says its autonomous security pipeline validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel in a single run. The run used an upcoming cyber model and cost $1.2 million in total API credits. The accompanying graphic shows 2,308 candidate findings narrowed to 1,724 validated findings and 948 working LPE exploits.
OpenAIOfficialOpenAI reports $1.2M AI vulnerability run cost about $1,279 per flawAIOpenAI says its AI vulnerability discovery run cost $1,213,430 in API credits, about $1,279 per distinct local privilege escalation vulnerability. It estimates that applying historical Pwn2Own and TyphoonPWN bounty rates would imply a hypothetical value of $18.9M to $66.3M. OpenAI argues the result shows AI models can make cyber attacks cost-prohibitive for attackers when defenders analyze their software proactively.
- OpenAIOfficialOpenAI describes a pipeline that finds, reproduces, and patches security vulnerabilities
AIOpenAI describes a pipeline that finds security-relevant code, scans for vulnerabilities, reproduces them in isolated environments, tests whether they can be exploited, and generates candidate patches. Human researchers review the results before reporting vulnerabilities and proposing fixes to maintainers.
- OpenAIOfficialPickOpenAI reports an autonomous pipeline that patched 948 Linux kernel privilege escalation bugs
AIOpenAI says its security researchers built an autonomous pipeline that discovers, exploits, and patches cyber vulnerabilities. In a single run with an upcoming cyber model, the pipeline validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel at a total API cost of $1.2 million. The accompanying funnel graphic shows 2,308 candidate findings narrowed to 1,724 validated findings, then 1,238 local privilege escalation candidates and 948 working exploits.
- OpenAI NewsOfficialPickOpenAI reports autonomous pipeline found and patched 948 Linux kernel vulnerabilities
AIOpenAI says it built an autonomous pipeline that found, validated, and generated patches for 948 local privilege-escalation vulnerabilities in the Linux kernel. The source is a brief excerpt and does not give further detail on methods or timing.