How AI Changes Vulnerability Discovery and Network Defense

MTS26m 37s
Sofia Puccini and Anand Oswal in blue tops flank the headline AI SECURITY RESET on black.
0 comments · 0 votesOpen discussionClose discussion
Sign in to join the discussion

    Palo Alto Networks executive Anand Oswal describes how frontier models can find software vulnerabilitiesA software vulnerability is a weakness in software that can permit an unintended or unauthorized effect on a system or its information. Its practical impact depends on the conditions under which it can be triggered. and potentially combine them into attack chains. He says his team's scan of more than 4,000 open-source packages found over 14,000 issues in two months, most not publicly known. These are the company's reported findings; the transcript does not provide an independent audit or establish that each item was exploited in the wild.

    The interview explains why ordinary software patching may lag rapid discovery. Some critical systems require lengthy testing or cannot be updated at all, so virtual patchingVirtual patching uses a separate security control to protect software while its underlying flaw remains. and network controls can serve as interim defenses. The guest claims that new protections can be prepared within hours, but that performance figure is a vendor assertion, and compensating controls do not remove the underlying flaw.

    The speakers also discuss AI agentsAn AI agent is a system that observes context, decides what to do, and takes actions through tools to pursue a goal. for security configuration, threat review and remediation. The proposed workflow begins with human approvalHuman-in-the-loop command approval requires a person to review and authorize a consequential AI action before it executes. and allows more autonomy only after repeated successful tasks, while retaining checks for false positivesA false positive is a result that incorrectly identifies a condition or category as present, such as flagging an irrelevant document as relevant.. Later examples of autonomous attacks are explicitly described as simulations with guardrails removed, not observed real-world incidents. A separate certificate and quantum-computing discussion adds context about cryptographic change but does not establish a timetable for a practical quantum threat.

    Original YouTube thumbnailWatch on YouTube

    Video topic coverage

    • Agents & workflows55
    • Coding & tools20
    • Models & learning25
    • Infrastructure & compute20
    • Research & evaluation35
    • Safety & security100
    • Business & productivity20
    • Creative & media0
    • Robotics & perception0
    • Society & policy10

    Share this video

    Sofia Puccini and Anand Oswal in blue tops flank the headline AI SECURITY RESET on black. With WWW.ARTIFICIAL-INTELLIGENCE.VIDEO, 6 August 2026 and duration 26m 37s.

    Anand Oswal says AI can accelerate both vulnerability discovery and attacks, making timely protections, context-aware defense and human review more important.