Summary
In this OpenAI conversation, a Standard Chartered security leader explains how the global bank is using frontier AI to protect its applications. As one of the first banks to evaluate OpenAI’s Daybreak, the team wanted to see whether frontier models could surface risks that traditional approaches miss across repositories, architectures and trust boundaries.
A central lesson is that context matters more than which model is chosen. Feeding the system architecture diagrams, deployment models, dependencies, connected services and the intent of each solution yields more findings that are truly exploitable, rather than a longer list of CVEs. The speaker stresses that attackers look at attack paths across systems, APIs, identities and trust boundaries, not just repos and assets.
The discussion also covers organizational change. AI-generated findings must be backed by hard evidence and validated against code, architecture and environment through cross-functional work between application security and development teams. Standard Chartered aims to move security from gatekeeper to enabler by shifting left, giving developers a test environment where they can see evidence of exploitability, and using frontier AI for remediation rather than adding another dashboard or backlog item. The interview closes with advice to get security foundations right and reduce technical debt.
📺 Source: OpenAI · Published September 29, 2026
🏷️ Format: Interview
