Newsletter Subscribe
Enter your email address below and subscribe to our newsletter
[forminator_form id="25163"]

forbes.com.openai.com+1.cryptobriefing.com+1.OpenAI says an autonomous research pipeline built 948 working local privilege-escalation exploits for the Linux kernel in a single run in mid-October, and wrote a candidate patch for each underlying bug. The total API cost was $1,213,430, or about $1,279 per exploit, according to the company's report, "A thousand fewer paths to root." OpenAI published that report on November 4. The findings are back in the news this week because new reporting shows the same model finding flaws in Apple's products.openai.com
OpenAI said its agents first narrowed the kernel to 5,034 audit targets that low-privileged users can reach. That came to about 2.8 million lines, or roughly 7% of the codebase. The agents then produced 2,308 candidate findings within 12 hours and reproduced 1,724 of them within 30 hours. The agents built 1,169 working root exploits and 291 information-leak proofs of concept. OpenAI says the 948 figure is the number of distinct patches written for the bugs behind those exploits. The exploits were tested on a stock desktop distribution with standard protections such as KASLR, SMEP and SMAP turned on. Human researchers reviewed the results before anything went to kernel maintainers.securityweek.com+2
The model behind the run was unnamed at first. OpenAI then released it to vetted users as GPT-6 Cyber on November 5.forbes.com
When the ten-day scan and review ended, other contributors had patched 61 of the bugs upstream and 43 had staging patches. OpenAI has not published CVE identifiers, and most fixes were still being reported. No major breaches have been tied to the findings so far, according to Crypto Briefing. OpenAI warned that, for now, a low-privileged account on a system that isn't fully patched "should be considered functionally equivalent to root access for a well-resourced threat actor." It recommended sandboxes such as Firecracker and gVisor to limit the damage.cryptobriefing.com+3
On Tuesday, Forbes reported that OpenAI and security firm Calif used GPT-6 Cyber to find about 35 weaknesses across Apple hardware and software. These include a way around the protections in Lockdown Mode. Apple's own security notes credit OpenAI Codex Security with three WebKit fixes, so far fewer than 35 are publicly attributed to OpenAI.cryptobriefing.com+1
Calif CEO Thai Duong told Forbes the outlook will get worse before it improves. "Between now and then, I think there will be more bugs, more exploits, and not just for the iPhone, but for the internet in general," he said.forbes.com