Microsoft, Palo Alto Networks Find Many Vulnerabilities by Using AI on Their Own Code

Microsoft and Palo Alto Networks have separately reported this week that they have seen significant results after turning AI on their own code to find vulnerabilities. Advanced AI models such as Claude Mythos have sparked debate in the cybersecurity industry about what the vulnerability discovery landscape will look like going forward. While some organizations have confirmed that these AI models are a game-changer, others are skeptical of their actual performance. Microsoft said on Tuesday that more than a dozen of the 137 vulnerabilities fixed with its latest Patch Tuesday updates were found by a new AI system called MDASH (multi-model agentic scanning harness) built by its Autonomous Code Security team. Palo Alto Networks revealed on Wednesday that it has used Claude Mythos and other frontier AI models to conduct a deep scan of its product portfolio, which resulted in the discovery of dozens of vulnerabilities.  Microsoft MDASH finds 16 vulnerabilities Microsoft’s MDASH system, which orchestrates more than 100 specialized AI agents across multiple frontier and distilled AI models, has been used to find vulnerabilities in the tech giant’s own codebases.  MDASH is designed to run a structured pipeline that moves findings through several distinct stages: preparation, scanning, validation, deduplication, and proof construction. Different agents handle different roles: some identify candidate vulnerabilities, others argue for or against their exploitability, and a final stage attempts to construct inputs that actually trigger the bug. This multi-stage debate architecture means that a finding must withstand scrutiny before it reaches a human engineer. According to Microsoft, MDASH was used to discover 16 of the vulnerabilities fixed with the latest Patch Tuesday updates. Four of them were rated critical, including unauthenticated remote code execution flaws in components such as the Windows kernel TCP/IP stack and the IKEv2 service.  Microsoft also tested MDASH against pre-patch snapshots of two heavily audited Windows components, and the AI recovered 96% and 100% of the confirmed vulnerabilities found over the past five years. In addition, on the public CyberGym benchmark (which includes 1,507 real-world vulnerability tasks), the AI system achieved an 88% rating. MDASH is currently in limited private preview, with Microsoft inviting security teams to apply for early access. Palo Alto Networks patches 75 vulnerabilities Palo Alto Networks typically publishes 5-10 advisories per month. However, on Wednesday it published 26 new advisories, a record credited to its early access to frontier AI models such as Mythos. The company used AI to analyze more than 130 products across SaaS-delivered and customer-operated environments, including products obtained via the recent acquisitions of CyberArk, Chronosphere, and Koi.  The 26 new advisories cover 75 vulnerabilities. While some were attributed to external researchers, the majority were detected internally using AI.  The cybersecurity giant pointed out that none of the 75 vulnerabilities are critical and there is no indication that they have been exploited in the wild. Three high-severity vulnerabilities were detected, but their exploitation requires highly specific configurations to be weaponized. Palo Alto Networks said it anticipates a surge in vulnerability discovery and patching as AI scanning becomes more widespread. The security firm believes organizations should act with urgency, as they have only a 3-5-month window to outpace adversaries.  The company noted that while its immediate priority is remediation of vulnerabilities, the long-term shift involves incorporating AI models directly into the software development lifecycle to prevent flaws from reaching production code. “Releasing 26 security advisories in a single day is a direct result of our internal security research utilizing Frontier AI models,” said Palo Alto Networks CISO Marc Benoit. “Volume does not equal severity; rather, it reflects our commitment to finding issues while their exploitation status remains ‘none known.’” Related: Google Detects First AI-Generated Zero-Day Exploit Related: Vulnerability in Claude Extension for Chrome Exposes AI Agent to Takeover Related: Cloudflare Lays Off 1,100 Employees in AI-Driven Restructuring WRITTEN BY Eduard Kovacs Eduard Kovacs (@EduardKovacs) is senior managing editor at SecurityWeek. He worked as a high school IT teacher before starting a career in journalism in 2011. Eduard holds a bachelor’s degree in industrial informatics and a master’s degree in computer techniques applied in electrical engineering. More from Eduard Kovacs Hundreds of Malicious Packages Force RubyGems to Suspend Registrations ICS Patch Tuesday: New Security Advisories From Siemens, Schneider, CISA BWH Hotels Says Hackers Had Access to Reservation Data for 6 Months Apple Patches Dozens of Vulnerabilities in macOS, iOS Claude Mythos Finds Only One Curl Vulnerability; Experts Divided on What It Really Means Frame Security Emerges From Stealth With $50M for Awareness and Training Platform Google Detects First AI-Generated Zero-Day Exploit Cloudflare Lays Off 1,100 Employees in AI-Driven Restructuring Latest News Foxconn Confirms North American Factories Hit by Cyberattack Sweet Security Launches Agentic AI Red Teaming to Counter ‘Mythos Moment’ Webinar Today: ROI for Cyber-Physical Security Programs Government to Scrutinize Instructure Over Canvas Disruption, Data Breach 716,000 Impacted by OpenLoop Health Data Breach Microsoft Patches Critical Zero-Click Outlook Vulnerability Threatening Enterprises Fortinet, Ivanti Patch Critical Vulnerabilities Chipmaker Patch Tuesday: Intel and AMD Patch 70 Vulnerabilities Trending Webinar: ROSI For CPS Security Programs May 13, 2026 In cyber-physical systems (CPS), just one hour of downtime can outweigh an entire annual security budget. Learn how to master the Return on Security Investment (ROSI) to align security goals with the bottom-line priorities. Register Virtual Event: Threat Detection And Incident Response Summit May 20, 2026 Delve into big-picture strategies to reduce attack surfaces, improve patch management, conduct post-incident forensics, and tools and tricks needed in a modern organization. Register People on the Move Malwarebytes has named Chung Ip as Chief Financial Officer. Semperis has appointed John Podboy as Chief Information Security Officer. Randy Menon has become Chief Product and Marketing Officer at One Identity. More People On The Move Expert Insights Is The SOC Obsolete, And We Just Haven’t Admitted It Yet? Many AI-first enterprises have already embraced sovereign architectures for general AI initiatives; cybersecurity—and the SOC—should be next. (Danelle Au) The Mythos Moment: Enterprises Must Fight Agents With Agents Only with the right platform and an agentic, AI-driven defense, will enterprises be able to protect themselves in the agentic era. (Etay Maor) Why Cybersecurity Must Rethink Defense In The Age Of Autonomous Agents From autonomous code generation to decision-making systems that initiate actions without human intervention, the industry is entering a new phase. (Torsten George) Government Can’t Win The Cyber War Without The Private Sector Securing national resilience now depends on faster, deeper partnerships with the private sector. (Steve Durbin) The Hidden ROI Of Visibility: Better Decisions, Better Behavior, Better Security Beyond monitoring and compliance, visibility acts as a powerful deterrent, shaping user behavior, improving collaboration, and enabling more accurate, data-driven security decisions. (Joshua Goldfarb) Flipboard Reddit Whatsapp Email