I u úspěchu AI je dobré vědět, jak k němu došla. Podle OpenAI v raném testu jejich systému získalo 20 z 200 úspěšných pokusů plná oprávnění v Linuxu přes jinou než zadanou chybu. xitter-newx.xyz-mikolabs.workers.dev/OpenAI/status/21181194…
Our goal is to help defenders find and fix vulnerabilities before attackers can use them. We’re working with Linux maintainers and Patch the Planet partners on reports and patches.
openai.com/index/a-thousand-…
1
3
I mean, shit, that's kind of something I can get behind.
A few less vuln in linux don't hurt.
Our goal is to help defenders find and fix vulnerabilities before attackers can use them. We’re working with Linux maintainers and Patch the Planet partners on reports and patches.
openai.com/index/a-thousand-…
2
80
Labs describe vulnerability detection and patching work.
OpenAI said it is working with Linux maintainers and Patch the Planet partners on vulnerability reports and patches.
Our goal is to help defenders find and fix vulnerabilities before attackers can use them. We’re working with Linux maintainers and Patch the Planet partners on reports and patches.
openai.com/index/a-thousand-…
1
8
Security: Report from OpenAI
@OpenAI says its AI pipeline found 948 distinct Linux kernel flaws enabling local root access, at $1.2M in API costs.
→ It reproduced 1,724 vulnerabilities before exploit testing
→ An early run used the wrong flaw in 20 of 200 successes, prompting independent exploit checks
Generated patches require maintainer review.
openai.com/index/a-thousand-…
Our goal is to help defenders find and fix vulnerabilities before attackers can use them. We’re working with Linux maintainers and Patch the Planet partners on reports and patches.
openai.com/index/a-thousand-…
37
Our goal is to help defenders find and fix vulnerabilities before attackers can use them. We’re working with Linux maintainers and Patch the Planet partners on reports and patches.
openai.com/index/a-thousand-…
5