CxOTalk is a long-form conversation series that explores technology strategy with the executives and innovators who are shaping it. Stuart McClure's appearance addresses adversarial AI in cybersecurity — a topic that was at the edge of the mainstream conversation when the episode was recorded and has since become one of the central challenges of the field.
The core of the conversation is a dual-use problem that Stuart has been articulating since Cylance's early days: the same machine learning capabilities that enable defenders to detect and predict malicious behavior at scale can be used by attackers to make malware more evasive, to automate attack discovery, and to generate credible social engineering content at volumes that were previously impossible. Understanding adversarial AI is not just a theoretical exercise — it is essential preparation for the attack landscape that is now arriving.
Stuart walks through how Cylance's prevention-first model addresses some of the adversarial AI challenge. A model that classifies executables based on their mathematical properties rather than known-bad signatures is inherently more resilient against adversarial manipulation than signature-based systems — though not immune. The conversation is candid about both the capabilities and the limitations of AI-based defense.
The CxOTalk format serves Stuart well. The extended conversation allows for the nuance and specificity that makes his thinking valuable to decision-makers — not the simplified talking points of a short media appearance, but the substantive analysis that helps executives understand what they are actually dealing with and what to do about it.