AI · Article
OpenAI System Finds 948 Privilege Escalation Flaws in Linux Kernel
Unite.AI ·

At a glance
An automated pipeline using OpenAI's cybersecurity model in development reported that it found 948 privilege escalation vulnerabilities in Linux and even created working attack code. Privilege escalation means a regular user gains administrator (root) access. A candidate fix was also created for each vulnerability found.
Key points
- The tests were run on virtual machines with default settings, without deliberately weakening security.
- After automatic analysis, 1,169 working attack codes were created. The fixes are only candidates and must be reviewed by people before use.
- In early experiments, there were 20 failed cases in which root access was gained through a vulnerability other than the target.
- The test API usage cost about $1.213 million, or roughly $1,279 per vulnerability.
Why it matters
Linux is widely used in Android devices and cloud servers, so finding and fixing these flaws quickly helps protect the devices and data of many people.
Original source
OpenAI 파이프라인, 리눅스 커널에서 권한 상승 취약점 948건 발견
Unite.AI · Nov 5, 09:15 KST
Read on Unite.AI ↗AI summarized the original article. Check the original for accuracy. Copyright belongs to Unite.AI.
OpenAI · Ranking and more news
More from this field
- AWS Quick Adds Chat-Built Flows, Updated Web App, and Teamwork Among Agents
AWS Quick added updates that let a team turn one person's work routine into a shared, repeatable process. Flows can now be built and changed entirely in chat, and they can run from the web, desktop, or mobile app. Agents working together can also give more complete answers.
Amazon AWS Machine Learning blog · Nov 21, 10:06 KST - Amazon Bedrock Web Search Adds Five Languages and Country-Based Results
Amazon's Web Search tool for Bedrock, which gives AI agents cited, up-to-date web results, now supports Spanish, French, German, Italian, and Japanese, plus non-US English. A new optional country field can also steer results toward sources from a specific country. Existing customers get the language update automatically, and pricing stays the same in every language.
AWS What's New · Nov 21, 09:17 KST - AWS Adds Shared Cache Skill to Cut Self-Hosted LLM Cost and Latency
AWS wrote about a shared cache that lets self-hosted AI models reuse saved work across servers, aiming to cut cost and wait time. The post shows how a coding agent can help set it up. Amazon says it can cut inference cost by up to 40% and first-token delay by up to 80%.
Amazon AWS Machine Learning blog · Nov 21, 09:14 KST - Can AI Really Find Black Friday TV Deals? A BBC Test Says Not Yet
A BBC team tested four popular AI chatbots to see if they could find a good Black Friday deal on a 55-inch TV. The results were mixed, with many links broken or out of date.
BBC News · Nov 21, 09:06 KST