Tag

Ai Vulnerability Discovery

All articles tagged with #ai vulnerability discovery

technology5 days ago•0 min saved

OpenAI’s GPT-6 Cyber Automates 948 Linux Exploits in 30-Hour Run

OpenAI has unveiled GPT-6 Cyber, a new AI model classified as 'Critical' under its Preparedness Framework, which identified 81 high-severity Firefox vulnerabilities and generated 948 working local privilege escalation exploits for the Linux kernel. The model achieved a 48.5% success rate on the ExploitGym benchmark, surpassing its predecessor. Access is currently restricted to approved customers in OpenAI’s Daybreak Red program, with plans to expand availability soon.

via BleepingComputer|

#ai-vulnerability-discovery #cybersecurity-defense #gpt-6-cyber

technology8 days ago•0 min saved

OpenAI’s AI Pipeline Identifies 948 Linux Kernel Exploits in 30-Hour Run

OpenAI demonstrated an autonomous AI pipeline that identified 948 local privilege escalation vulnerabilities in the Linux kernel within 30 hours. The process, which cost $1.2 million in API credits, generated working exploits and candidate patches for human review. While the cost is significantly lower than traditional bug bounties, experts warn that such AI capabilities could accelerate both defensive patching and offensive exploitation by adversaries.

via Phoronix|

#ai-vulnerability-discovery #cybersecurity #linux-kernel

security3 months ago•3 min saved

North Korea Exploits Windows Zero-Day in New 'Dream Job' Campaign Amid 421-Patch Surge

Microsoft patched 421 vulnerabilities in its August 2026 release, including a zero-day in the Windows Ancillary Function Driver (afd.sys) actively exploited by North Korea’s Lazarus Group. The flaw, CVE-2026-68820, allows local privilege escalation to SYSTEM level and was used in the 'Operation Dream Job' campaign targeting defense sectors in Europe and India. While the patch volume is high due to AI-assisted discovery, experts warn that only one bug is confirmed exploited, urging prioritized rather than rushed deployment.

via The Register|

#ai-vulnerability-discovery #cve-2026-68820 #lazarus-group

cybersecurity3 months ago•2 min saved

Microsoft Patches 398 Flaws, Including Zero-Day Exploited by North Korea's Lazarus Group

Microsoft released its August 2026 Patch Tuesday updates, addressing 398 vulnerabilities across Windows and other software. The release includes a critical zero-day, CVE-2026-68820, which is under active exploitation by North Korea's Lazarus Group. While AI has accelerated the discovery of these flaws, experts warn that automated patching remains unreliable, requiring human oversight to ensure fixes do not introduce new weaknesses.

via Krebs on Security|

#ai-vulnerability-discovery #cve-2026-68820 #cybersecurity