• Design and implement security controls in the inference gateways, platforms and developer tooling used across the firm, delivered as code and configuration.
• Embed controls in the systems that engineering teams already work in, so that the secure option is the default.
• Define the standards and patterns other teams adopt when building with AI.
• Define who an internal agent acts as, what it is permitted to reach and how those decisions are enforced.
• Build telemetry that records what ran, on whose behalf and against what.
• Establish how agent and automation activity is proven after the fact for audit and investigation.
• Secure developer harnesses, plugins, and the dependency and model provenance they rely on.
• Own secrets handling and vulnerability management across the AI tooling the firm depends on.
• Set provenance and integrity requirements for models and components introduced into the firm.
• Threat model live systems across platform and gateway security, agent identity and permissioning, and the AI supply chain.
• Deliver the remediation resulting from threat modelling directly.
• Identify gaps in control coverage and prioritise remediation against risk and delivery constraints.
• Take and document control decisions in areas where no established answer exists.
• Set and own the technical direction for the function’s core areas and help define how the function works.
• Drive the adoption of secure defaults and governed identities and gateways with a reliable audit trail.