OpenAI on Wednesday detailed a pipeline that employed an upcoming cyber-focused artificial intelligence (AI) model to validate and patch 948 local privilege escalation vulnerabilities in Linux for $1.2 million in API costs.
“Our prototype identified and analyzed 2,308 candidate Linux kernel findings, dynamically reproduced 1,724 vulnerabilities, and produced 948 local privilege escalation (LPE) exploits that could allow an unprivileged local user to gain root access,” the AI company said.
The findings build upon 24 LPE exploits that OpenAI revealed were discovered by its GPT‑5.5‑Cyber model back in June 2026. The latest research turns this process into an agentic pipeline comprising five stages –
- Vulnerability discovery, where agents identify high-value audit targets (i.e., parts of the codebase that are security-relevant and reachable by low-privileged users) and look for candidate findings
- Dynamic validation, where agents attempt to reproduce each candidate in a sandboxed environment and preserve runtime evidence
- Exploitability assessment, where agents test whether validated vulnerabilities could be abused to facilitate information disclosure or LPE and devise a proof-of-concept (PoC) exploit
- Patch generation, where agents generate patches for the identified vulnerability
- Human review, where security researchers review the exploit, patch, and proposed severity before it’s reported to the maintainers
“In a run in mid-October 2026, our pipeline using an upcoming cyber model dynamically reproduced 1,724 vulnerabilities [within 30 hours] and generated PoCs for 291 information leak vulnerabilities and 948 working local privilege escalation (LPE) exploits for a commonly used Linux distribution,” OpenAI said.
“The total run cost $1,213,430 in API credits, or $1,279 per LPE exploit. To the best of our knowledge, the agents in this pipeline discovered and used at least 10 novel Linux exploitation techniques.”
At a high level, these include –
- Two new persistent elastic objects to spray objects inside the target kernel cache
- Two new page grooming techniques to improve exploit reliability
- Five new data-only techniques to obtain privilege escalation directly or via an arbitrary write
- One technique that involves arbitrary type confusion
The development is once again an indication of how frontier AI is redrawing the contours of cybersecurity, accelerating vulnerability discovery and exploitation, and at the same time helping defenders reliably address defects, albeit through additional instructions and iterations, before attackers can use them, and make it “cost prohibitive” for bad actors to exploit software.
In a report published last month, VulnCheck categorized AI-assisted vulnerability discovery as a “growing risk category” that turbocharges novice hackers, who now require little to no technical expertise to detect software vulnerabilities.
The access to AI models is not just limited to defenders, but also bad actors who can leverage the same capabilities to discover and weaponize new zero-days and expand the pool of viable targets with minimal effort.
According to insights from Google Threat Intelligence Group (GTIG), the number of vulnerabilities disclosed per month has more than doubled, rising from 5,045 in January 2026 to 10,740 in August 2026. The number of vulnerabilities classified as high-risk has surged from 131 disclosures in January 2026 to 350 in August 2026, a 167% growth.
“The widespread availability of AI-powered cyber tools has almost certainly led to a significant increase in the discovery of zero-day vulnerabilities,” the U.K. National Cyber Security Centre (NCSC) said in an assessment published Tuesday.
“Defensive AI-cyber tools have reduced the window of a vulnerability moving from zero-day to an N-day vulnerability, and have made patching a quicker process. However, without correct and widespread configuring of this technology, many systems are still vulnerable to attackers using AI-cyber tools to identify and exploit system weaknesses.”
The NCSC also pointed out that organizations will have to evolve their internal patching and testing processes to meet this new pace of vulnerability discovery.
“Attackers have subsequently displayed less caution over losing an access to achieve an objective, as the cost to replace is minimal,” it added. “Cyber capabilities are now used for a broader range of activity due to the lower risk involved, making it easier for adversaries to access information and use it operationally.”
OpenAI said the rate of vulnerability discovery by frontier models needs to be matched by automated patch generation to secure software and to limit the exploitation window.
“With human expert-guided iteration, we believe AI-generated patches can be functionally correct, aligned with existing coding styles, and robust to attack variants,” OpenAI said. “For the near and medium term, having access to a low-privileged account on an operating system that isn’t up-to-date on its latest available patches should be considered functionally equivalent to root access for a well-resourced threat actor.”
